WO2013081051A1 - Recommendation device, recommendation system, recommendation method and program - Google Patents

Recommendation device, recommendation system, recommendation method and program Download PDF

Info

Publication number
WO2013081051A1
WO2013081051A1 PCT/JP2012/080921 JP2012080921W WO2013081051A1 WO 2013081051 A1 WO2013081051 A1 WO 2013081051A1 JP 2012080921 W JP2012080921 W JP 2012080921W WO 2013081051 A1 WO2013081051 A1 WO 2013081051A1
Authority
WO
WIPO (PCT)
Prior art keywords
recommendation
history information
similarity
user
analysis processing
Prior art date
Application number
PCT/JP2012/080921
Other languages
French (fr)
Japanese (ja)
Inventor
祥 佐々木
Original Assignee
Kddi株式会社
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Kddi株式会社 filed Critical Kddi株式会社
Priority to CN201280058706.2A priority Critical patent/CN104246751B/en
Publication of WO2013081051A1 publication Critical patent/WO2013081051A1/en

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/90—Details of database functions independent of the retrieved data types
    • G06F16/95—Retrieval from the web
    • G06F16/953—Querying, e.g. by the use of web search engines
    • G06F16/9535—Search customisation based on user profiles and personalisation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q30/00—Commerce
    • G06Q30/02—Marketing; Price estimation or determination; Fundraising
    • G06Q30/0201—Market modelling; Market analysis; Collecting market data
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q30/00—Commerce
    • G06Q30/02—Marketing; Price estimation or determination; Fundraising
    • G06Q30/0282—Rating or review of business operators or products
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q30/00—Commerce
    • G06Q30/06—Buying, selling or leasing transactions
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q30/00—Commerce
    • G06Q30/06—Buying, selling or leasing transactions
    • G06Q30/0601—Electronic shopping [e-shopping]
    • G06Q30/0631—Item recommendations

Definitions

  • the present invention relates to a recommendation device, a recommendation system, a recommendation method, and a program that perform a long-term preference analysis process and a short-term interest analysis process, integrate the recommendation results of both processes, and output a final recommendation result.
  • Non-Patent Document 1 automated collaborative filtering is proposed.
  • Non-Patent Document 2 etc.
  • collaborative filtering based on a probability model is proposed, while in Non-Patent Document 3, non-patent An attempt was made to speed up the collaborative filtering of Document 1.
  • Non-Patent Document 4 proposes a collaborative filtering technique capable of placing importance on the latest history information by assuming a preference decay model in the user history information.
  • Non-Patent Document 1 and Non-Patent Document 2 since all the history information that has been accumulated is analyzed and a recommendation result is generated, if the history is accumulated more than necessary by the user, the recommendation The content calculated by the processing is fixed each time, and there is a problem that new content cannot be found.
  • Non-Patent Document 3 since the similarity is defined only between the representative user and the user to be recommended, the amount of calculation to be analyzed at the time of recommendation compared with the technique described in Non-Patent Document 1
  • Non-Patent Document 3 focuses only on speeding up the recommendation process and considers the latest history information. However, if the history is accumulated more than necessary by the user as described above, there is still a problem that the content presented by the recommendation is fixed every time. .
  • Non-Patent Document 4 emphasis is placed on the latest history information so as to allow a change in the user's preference, but it does not take into account the momentary excitement of the user, and is more up-to-date. It is intended to analyze the preference based on the history information. Therefore, there is a problem that followability to a recommendation with respect to a one-time interest that is out of the long-term preference of the user is low.
  • the present invention has been made in view of the above-described problems, and a recommendation device, a recommendation system, which achieves a highly accurate recommendation that accurately corresponds to what the user is interested in in the short term, It is an object to provide a recommendation method and program.
  • the present invention proposes the following items in order to solve the above problems.
  • symbol corresponding to embodiment of this invention is attached
  • the present invention defines similarity between users based on user history information such as purchase history and browsing history, and recommends content that is predicted to be of interest to the user based on the defined similarity.
  • a recommendation device that uses the collaborative filtering method to output as a result, which analyzes all the history information and generates information necessary for the recommendation (for example, the long-term preference analysis processing unit in FIG. 1) 1100), short-term interest analysis processing means (for example, equivalent to the short-term interest analysis processing unit 1200 in FIG.
  • the recommendation apparatus 1) for analyzing history information for each session and generating information necessary for recommendation, and the long-term Recommendation result output means (for example, output of recommendation results by integrating the output of the preference analysis processing means and / or the output of the short-term interest analysis processing means (for example, 1 corresponding to the recommendation result output unit 1300), and an integration ratio calculation unit that changes the recommendation result of the recommendation result output unit based on the transition of the user's interest (for example, corresponding to the integration ratio calculation unit 1600 in FIG. 1);
  • the recommendation apparatus characterized by having provided these.
  • the long-term preference analysis processing means analyzes all history information and generates information necessary for the recommendation.
  • the short-term interest analysis processing means analyzes history information for each session and generates information necessary for recommendation.
  • the recommendation result output means integrates the output of the long-term preference analysis processing means and / or the output of the short-term interest analysis processing means, and outputs a recommendation result.
  • the integration ratio calculation unit changes the recommendation result of the recommendation result output unit based on the transition of the user's interest. Therefore, by providing a recommendation result based on the long-term preference of the user by the long-term preference analysis processing means, and by providing a recommendation result for the short-term interest in the situation by the short-term interest analysis processing means, It is possible to provide an adaptive recommendation result considering the transition of interest.
  • the long-term preference analysis processing means collects history information necessary for analysis from all log information (for example, first history information collection means (for example, 2) and first similarity calculation means for analyzing the collected history information and defining the similarity between users (for example, the similarity calculation unit 1102 in FIG. 2).
  • First recommendation degree calculating means for calculating the recommendation degree of each content for the user using the calculated similarity and the history information necessary for the analysis collected from all the log information (For example, equivalent to the recommendation degree calculation unit 1103 in FIG. 2) and storage means for storing the calculated recommendation degree (for example, equivalent to the storage part 1104 in FIG.
  • the second history information collection means (for example, equivalent to the history information collection unit 1201 in FIG. 2) that collects the history information necessary for analysis from among the users, and analyzes the collected history information
  • Second similarity calculation means for defining similarity for example, equivalent to the similarity calculation unit 1202 in FIG. 2
  • a second recommendation level calculation unit for example, equivalent to the recommendation level calculation unit 1203 in FIG.
  • the recommendation result output unit includes the integration According to the change result of the ratio calculation means, a recommendation result is output from the stored recommendation degree for each content and the recommendation degree for each content calculated by the second recommendation degree calculation means, It has proposed that recommendation apparatus.
  • the first history information collection means of the long-term preference analysis processing means collects history information necessary for analysis from all log information.
  • the first similarity calculation unit analyzes the collected history information and defines the similarity between users.
  • the first recommendation degree calculating means calculates the recommendation degree of each content for the user using the calculated similarity and the history information necessary for the analysis collected from all the log information.
  • the storage means stores the calculated recommendation level.
  • the second history information collection means of the short-term interest analysis processing means collects history information necessary for analysis from the log information for each session in real time.
  • the second similarity calculation means analyzes the collected history information and defines the similarity between users.
  • the second recommendation degree calculation means calculates the recommendation degree of each content for the user using the calculated similarity and history information necessary for analysis from the log information for each session.
  • the recommendation result output unit outputs a recommendation result from the stored recommendation level for each content and the recommendation level for each content calculated by the second recommendation level calculation unit according to the change result of the integration ratio calculation unit. Therefore, by providing a recommendation result based on the long-term preference of the user by the long-term preference analysis processing means, and by providing a recommendation result for the short-term interest in the situation by the short-term interest analysis processing means, It is possible to provide an adaptive recommendation result considering the transition of interest.
  • the long-term preference analysis processing means collects history information necessary for analysis from all log information (for example, first history information collection means (for example, 5) and a first similarity calculation unit that analyzes the collected history information and defines the similarity between users (for example, the similarity calculation unit 1102 in FIG. 5). And a storage means (for example, equivalent to the storage unit 1104 in FIG. 5) for storing the calculated similarity, and the short-term interest analysis processing means analyzes the log information for each session.
  • Second history information collection means for example, equivalent to the history information collection unit 1201 in FIG.
  • Second similarity calculator (For example, equivalent to the similarity calculation unit 1202 in FIG. 5) and the history information necessary for analysis from the log information for each session and the stored similarity, the degree of recommendation of each content to the user From the first recommendation degree calculation means (for example, equivalent to the recommendation degree calculation unit 1203 in FIG. 5), the similarity calculated by the second similarity calculation means, and the log information for each session Second recommendation level calculation means (for example, equivalent to the recommendation level calculation unit 1215 in FIG.
  • the output unit calculates the recommendation level for each content calculated by the first recommendation level calculation unit and the content calculated by the second recommendation level calculation unit according to the change result of the integration ratio calculation unit. It proposes a recommendation apparatus and outputting a recommendation result from a recommendation degree with.
  • the first history information collection means of the long-term preference analysis processing means collects history information necessary for analysis from all log information.
  • the first similarity calculation unit analyzes the collected history information and defines the similarity between users.
  • the storage means stores the calculated similarity.
  • the second history information collection means of the short-term interest analysis processing means collects history information necessary for analysis from the log information for each session in real time.
  • the second similarity calculation means analyzes the collected history information and defines the similarity between users.
  • the first recommendation level calculation means calculates the recommendation level of each content for the user by using the history information necessary for the analysis and the stored similarity from the log information for each session.
  • the second recommendation level calculation means uses the similarity calculated by the second similarity calculation means and the history information necessary for analysis from the log information for each session to determine the recommendation level of each content to the user. calculate.
  • the recommendation result output means is based on the recommendation degree for each content calculated by the first recommendation degree calculation means and the recommendation degree for each content calculated by the second recommendation degree calculation means according to the change result of the integration ratio calculation means. Output recommendation results. Therefore, by providing a recommendation result based on the long-term preference of the user by the long-term preference analysis processing means, and by providing a recommendation result for the short-term interest in the situation by the short-term interest analysis processing means, It is possible to provide an adaptive recommendation result considering the transition of interest.
  • the long-term preference analysis processing means collects history information necessary for analysis from all log information (for example, first history information collection means (for example, 8) and first similarity calculation means for analyzing the collected history information and defining the similarity between users (for example, the similarity calculation unit 1102 in FIG. 8).
  • First recommendation degree calculation means for example, a degree of recommendation of each content for the user
  • Storage means for storing the defined similarity and the calculated recommendation degree.
  • the short-term interest analysis processing means includes log information for each session.
  • Second history information collecting means for collecting history information in real time for example, corresponding to the history information collecting unit 1201 in FIG. 8
  • Second recommendation degree calculation means for example, equivalent to the recommendation degree calculation unit 1226 in FIG. 8 that calculates the recommendation degree of each content for the user using history information necessary for the analysis
  • the recommendation result output means A recommendation result from the recommendation level for each content calculated by the first recommendation level calculation unit and the recommendation level for each content calculated by the second recommendation level calculation unit according to the change result of the integration ratio calculation unit.
  • the first history information collection means of the long-term preference analysis processing means collects history information necessary for analysis from all log information.
  • the first similarity calculation unit analyzes the collected history information and defines the similarity between users.
  • the first recommendation level calculating means calculates the recommendation level of each content for the user using the calculated similarity and history information necessary for analysis from the log information for each session.
  • the storage means stores the defined similarity and the calculated recommendation level.
  • the second history information collection means of the short-term interest analysis processing means collects history information necessary for analysis from the log information for each session in real time.
  • the second recommendation degree calculation means calculates the recommendation degree of each content for the user using the defined similarity, the calculated recommendation degree, and history information necessary for analysis from the log information for each session.
  • the recommendation result output means is based on the recommendation degree for each content calculated by the first recommendation degree calculation means and the recommendation degree for each content calculated by the second recommendation degree calculation means according to the change result of the integration ratio calculation means. Output recommendation results. Therefore, by providing a recommendation result based on the long-term preference of the user by the long-term preference analysis processing means, and by providing a recommendation result for the short-term interest in the situation by the short-term interest analysis processing means, It is possible to provide an adaptive recommendation result considering the transition of interest.
  • the present invention acquires the browsing status of the user with respect to the recommendation device of (1) to (4), with respect to the database storing the log information for each session and the recommendation output from the recommendation result output means.
  • the recommendation device is characterized in that the ratio calculation means changes the recommendation result of the recommendation result output means based on the transition of the user's interest based on the browsing status feedback information.
  • the database stores log information for each session.
  • the browsing status acquisition unit acquires the browsing status of the user for the recommendation output from the recommendation result output unit.
  • the browsing status feedback means feeds back the acquired browsing status of the user to the database and the integrated ratio calculation means.
  • the integrated ratio calculation means changes the recommendation result of the recommendation result output means based on the transition of the user's interest based on the browsing status feedback information. Therefore, the obtained browsing status of the user is fed back to the database and the integrated ratio calculating means, and the integrated ratio calculating means changes the recommended result of the recommended result output means based on the transition of the user's interest based on the browsing status feedback information. Therefore, it is possible to make a recommendation by accurately extracting contents and the like that the user is interested in in the short term.
  • the present invention defines similarity between users based on user history information such as purchase history and browsing history, and recommends content that is predicted to be of interest to the user based on the defined similarity.
  • a recommendation system that uses the collaborative filtering method that is output as a result, the first database storing all the history information (for example, corresponding to the database 200 in FIG. 10), and the first database storing the history information for each session.
  • a long-term preference analysis processing apparatus for example, FIG. 10) that generates history information stored in the first database and generates information necessary for recommendation. Long-term preference analysis processing device 400), and history information for each session stored in the second database.
  • a short-term interest analysis processing device for example, equivalent to the short-term interest analysis processing device 500 in FIG.
  • a recommendation result output device (for example, equivalent to the recommendation result output device 600 in FIG. 10) that integrates the output of the interest analysis processing device and outputs a recommendation result, and a recommendation result of the recommendation result output device as a transition of the user's interest
  • a recommendation system is proposed that includes an integration rate calculation device (for example, equivalent to the integration rate calculation device 900 in FIG. 10) that is changed based on this.
  • the first database stores all history information.
  • the second database stores history information for each session.
  • the long-term preference analysis processing device analyzes history information stored in the first database and generates information necessary for the recommendation.
  • the short-term interest analysis processing device analyzes the history information for each session stored in the second database and generates information necessary for the recommendation.
  • the recommendation result output device integrates the output of the long-term preference analysis processing device and / or the output of the short-term interest analysis processing device, and outputs a recommendation result.
  • the integrated ratio calculation device changes the recommendation result of the recommendation result output device based on the transition of the user's interest.
  • the long-term preference analysis processing device collects history information necessary for analysis from all log information (for example, first history information collection means (for example, 11) and first similarity calculation means for analyzing the collected history information and defining the similarity between users (for example, the similarity calculation unit 402 in FIG. 11).
  • First recommendation degree calculating means for calculating the recommendation degree of each content for the user using the calculated similarity and the history information necessary for the analysis collected from all the log information (For example, equivalent to the recommendation degree calculation unit 403 in FIG. 11) and storage means (for example, equivalent to the storage unit 404 in FIG.
  • a second history information collecting unit for example, corresponding to the history information collecting unit 501 in FIG. 11
  • a second similarity calculation means for example, equivalent to the similarity calculation unit 502 in FIG. 11
  • Second recommendation level calculation means for example, equivalent to the recommendation level calculation unit 503 in FIG. 11
  • a recommendation result is output from the stored recommendation level for each content and the recommendation level for each content calculated by the second recommendation level calculation means according to the change result of the integrated ratio calculation device. It has proposed a recommendation system that.
  • the first history information collecting means of the long-term preference analysis processing device collects history information necessary for analysis from all log information.
  • the first similarity calculation unit analyzes the collected history information and defines the similarity between users.
  • the first recommendation degree calculating means calculates the recommendation degree of each content for the user using the calculated similarity and the history information necessary for the analysis collected from all the log information.
  • the storage means stores the calculated recommendation level.
  • the second history information collecting means of the short-term interest analysis processing device collects history information necessary for analysis from the log information for each session in real time.
  • the second similarity calculation means analyzes the collected history information and defines the similarity between users.
  • the second recommendation degree calculation means calculates the recommendation degree of each content for the user using the calculated similarity and history information necessary for analysis from the log information for each session.
  • the recommendation result output device outputs a recommendation result from the stored recommendation level for each content and the recommendation level for each content calculated by the second recommendation level calculation means according to the change result of the integrated ratio calculation device. Therefore, by providing a recommendation result based on the long-term preference of the user by the long-term preference analysis processing device, and by providing a recommendation result for the short-term interest in the situation by the short-term interest analysis processing device, It is possible to provide an adaptive recommendation result considering the transition of interest.
  • the long-term preference analysis processing device collects history information necessary for analysis from all the log information (for example, first history information collection means (for example, 14) and first similarity calculation means for analyzing the collected history information and defining the similarity between users (for example, the similarity calculation unit 402 in FIG. 14). And a storage means (for example, corresponding to the storage unit 415 in FIG. 14) for storing the calculated similarity, and the short-term interest analysis processing device analyzes the log information for each session. Second history information collecting means for collecting necessary history information in real time (e.g., corresponding to the history information collection unit 501 in FIG.
  • Second similarity calculation Each content is recommended to the user using the level (for example, equivalent to the similarity calculation unit 502 in FIG. 14), the history information necessary for analysis from the log information for each session, and the stored similarity.
  • a first recommendation degree calculating means for calculating the degree for example, equivalent to the recommendation degree calculating unit 503 in FIG. 14
  • the similarity calculated by the second similarity calculating means and the log information for each session.
  • Second recommendation level calculation means (for example, equivalent to the recommendation level calculation unit 515 in FIG.
  • the result output device determines the recommendation level for each content calculated by the first recommendation level calculation unit and the content calculated by the second recommendation level calculation unit according to the change result of the integrated ratio calculation unit. It proposes a recommendation system and outputs the recommendation result from the recommendation degree of each.
  • the first history information collecting means of the long-term preference analysis processing device collects history information necessary for analysis from all log information.
  • the first similarity calculation unit analyzes the collected history information and defines the similarity between users.
  • the storage means stores the calculated similarity.
  • the second history information collecting means of the short-term interest analysis processing device collects history information necessary for analysis from the log information for each session in real time.
  • the second similarity calculation means analyzes the collected history information and defines the similarity between users.
  • the first recommendation level calculation means calculates the recommendation level of each content for the user by using the history information necessary for the analysis and the stored similarity from the log information for each session.
  • the second recommendation level calculation means uses the similarity calculated by the second similarity calculation means and the history information necessary for analysis from the log information for each session to determine the recommendation level of each content to the user. calculate.
  • the recommendation result output device uses the recommendation level for each content calculated by the first recommendation level calculation unit and the recommendation level for each content calculated by the second recommendation level calculation unit according to the change result of the integrated ratio calculation device. Output recommendation results. Therefore, by providing a recommendation result based on the long-term preference of the user by the long-term preference analysis processing means, and by providing a recommendation result for the short-term interest in the situation by the short-term interest analysis processing means, It is possible to provide an adaptive recommendation result considering the transition of interest.
  • the long-term preference analysis processing device collects history information necessary for analysis from all the log information (for example, first history information collection means (for example, 17) and first similarity calculation means for analyzing the collected history information and defining the similarity between users (for example, the similarity calculation unit 402 in FIG. 17).
  • First recommendation degree calculating means for example, calculating the recommendation degree of each content for the user, using the calculated similarity and the history information necessary for analysis from the log information for each session) 17
  • a storage means for example, corresponding to the storage unit 426 in FIG. 17 for storing the defined similarity and the calculated recommendation level.
  • the processing unit Second history information collection means for example, corresponding to the history information collection unit 501 in FIG.
  • the recommendation system is characterized in that the recommendation result is output from the recommendation level calculated for each content.
  • the first history information collecting means of the long-term preference analysis processing device collects history information necessary for analysis from all log information.
  • the first similarity calculation unit analyzes the collected history information and defines the similarity between users.
  • the first recommendation level calculating means calculates the recommendation level of each content for the user using the calculated similarity and history information necessary for analysis from the log information for each session.
  • the storage means stores the defined similarity and the calculated recommendation level.
  • the second history information collecting means of the short-term interest analysis processing device collects history information necessary for analysis from the log information for each session in real time.
  • the second recommendation degree calculation means calculates the recommendation degree of each content for the user using the defined similarity, the calculated recommendation degree, and history information necessary for analysis from the log information for each session.
  • the recommendation result output device uses the recommendation level for each content calculated by the first recommendation level calculation unit and the recommendation level for each content calculated by the second recommendation level calculation unit according to the change result of the integrated ratio calculation device. Output recommendation results. Therefore, by providing a recommendation result based on the long-term preference of the user by the long-term preference analysis processing device, and by providing a recommendation result for the short-term interest in the situation by the short-term interest analysis processing device, It is possible to provide an adaptive recommendation result considering the transition of interest.
  • the present invention relates to the recommendation system of (6) to (9), a browsing status acquisition apparatus that acquires the browsing status of the user for the recommendations to be output, and the acquired browsing status of the user in the second database. And a browsing status feedback device (for example, corresponding to the browsing status feedback device 800 of FIG. 10) that feeds back to the integrated ratio calculating device, and the integrated ratio calculating device outputs the recommendation result based on the browsing status feedback information
  • the recommendation system characterized by changing the recommendation result of an apparatus based on transition of a user's interest is proposed.
  • the browsing status acquisition device acquires the browsing status of the user for the recommendations output from the recommendation result output device.
  • the browsing status feedback device feeds back the acquired browsing status of the user to the database and the integrated ratio calculation device.
  • the integrated ratio calculation device changes the recommendation result of the recommendation result output device based on the transition of the user's interest based on the browsing status feedback information. Therefore, the obtained browsing status of the user is fed back to the database and the integrated rate calculating device, and the integrated rate calculating device changes the recommended result of the recommended result output device based on the transition of the user's interest based on the browsing status feedback information. Therefore, it is possible to make a recommendation by accurately extracting contents and the like that the user is interested in in the short term.
  • the present invention defines similarity between users based on user history information such as purchase history and browsing history, and recommends content predicted to be of interest to the user based on the defined similarity. It is a recommendation method using a collaborative filtering method that is output as a result, and a long-term preference analysis processing step that analyzes all the history information to generate information necessary for the recommendation, and analyzes the history information for each session to recommend A short-term interest analysis processing step for generating necessary information, and a recommendation result output step for outputting a recommendation result by integrating the output of the long-term preference analysis processing step and / or the output of the short-term interest analysis processing step; , Integration to change the recommendation result of the recommendation result output step based on the transition of the user's interest It proposes a recommendation method characterized by comprising: a case calculating step.
  • all the history information is analyzed to generate information necessary for the recommendation, and the history information for each session is analyzed to generate information necessary for the recommendation. Then, the above two processing outputs are integrated to output a recommendation result, and the recommendation result is changed based on the transition of the user's interest. Therefore, by providing a recommendation result based on the long-term preference of the user by the long-term preference analysis processing step, and by providing a recommendation result for the short-term interest in the situation by the short-term interest analysis processing step, It is possible to provide an adaptive recommendation result considering the transition of interest.
  • the present invention defines similarity between users based on user history information such as purchase history and browsing history, and recommends content predicted to be of interest to the user based on the defined similarity.
  • all the history information is analyzed to generate information necessary for the recommendation, and the history information for each session is analyzed to generate information necessary for the recommendation. Then, the above two processing outputs are integrated to output a recommendation result, and the recommendation result is changed based on the transition of the user's interest. Therefore, by providing a recommendation result based on the long-term preference of the user by the long-term preference analysis processing step, and by providing a recommendation result for the short-term interest in the situation by the short-term interest analysis processing step, It is possible to provide an adaptive recommendation result considering the transition of interest.
  • a recommendation result based on a user's long-term preference is provided by a long-term preference analysis process, and a recommendation result for a short-term interest in the situation is provided by a short-term interest analysis process.
  • the recommendation device 1000 includes a long-term preference analysis processing unit 1100, a short-term interest analysis processing unit 1200, a recommendation result output unit 1300, a recommendation display unit 1400, and a browsing situation.
  • the database 200 is connected to the long-term preference analysis processing unit 1100
  • the database 300 is connected to the short-term interest analysis processing unit 1200.
  • the database 200 stores all the log data of all users, for example, all the databases that record the purchase history of products and the browsing history of web pages.
  • the database 300 stores log data for each session in order to perform short-term interest analysis processing. Specifically, from the above database, for example, in the browsing history of web pages, each user's continuous browsing history within 30 minutes is subdivided into one session, or randomly selected from them. Or two web page sets that co-occur between each session, and the web page set that exceeds the separately set threshold is regarded as a related web page, and a plurality of related web pages are Create a database that considers history as browsing history.
  • the data reduction method is not limited to this method.
  • the long-term preference analysis processing unit 1100 acquires history information of the recommended user and all other users from all databases, outputs the similarity between the recommended user and other users as the similarity, The degree of recommendation of each content for the recommended user is output as a recommendation degree from the output of the above and the history information of all databases, and the output is stored. The above processing is performed for all users by changing the recommended user.
  • the short-term interest analysis processing unit 1200 from the contracted database, similarly to the above, a web page being continuously browsed within 30 minutes is set as a session, and the content included in the session is set as the user's content.
  • the similarity between the recommended user and other users is output as the similarity, and the recommended degree of the recommended content of each content for the recommended user from the above output and the history information of the reduced database Output as.
  • the recommendation result output unit 1300 acquires the recommendation level of each content for the recommended user from the record stored in the long-term preference analysis processing unit 1100.
  • the output from the short-term interest analysis processing unit 1200 is acquired at the same time, and the recommendation level of each content is finalized by a method determined by the integration ratio calculation unit 1600, for example, taking a weighted average of the two recommendation levels. The degree of recommendation is calculated.
  • the integration ratio calculation unit 1600 takes into account the number of contents of the user's session Num (session) and the browsing time Time (session), for example, the following mathematical formulas (1), (2), (3)
  • the recommendation level Rec (batch) of each content by the long-term preference analysis process and the recommendation level Rec (real) of each content by the short-term interest analysis process are integrated by combining them, and the final recommendation level Rec (Final) Is calculated.
  • the recommendation device includes a history information collection unit 1101, a similarity calculation unit 1102, a recommendation degree calculation unit 1103, and a storage unit 1104 that constitute a long-term preference analysis processing unit 1100.
  • an integrated ratio calculation unit 1600 is an integrated ratio calculation unit 1600.
  • the history information collection unit 1101 collects history information necessary for analysis from all the log information from the database 200.
  • the similarity calculation unit 1102 analyzes the collected history information and defines the similarity between users.
  • the recommendation level calculation unit 1103 calculates the recommendation level of each content for the user by using the calculated similarity and history information necessary for analysis collected from all log information.
  • the storage unit 1104 stores the calculated recommendation level.
  • the history information collection unit 1201 collects history information necessary for analysis from the log information for each session in real time in the database 300.
  • the similarity calculation unit 1202 analyzes the collected history information and defines the similarity between users.
  • the recommendation level calculation unit 1203 calculates the recommendation level of each content for the user using the calculated similarity and history information necessary for analysis from the log information for each session.
  • the recommendation result output unit 1300 outputs a recommendation result from the stored recommendation level for each content and the recommendation level for each content calculated by the recommendation level calculation unit 1203 according to the change result of the integration ratio calculation unit 1600.
  • the function of the integration ratio calculation unit 1600 is to change the weighted sum, and is processed by an index / logistic function or the like based on, for example, the number of sessions viewed or the session time.
  • the history information collection unit 1101 of the long-term preference analysis processing unit 1100 collects history information necessary for analysis from all the log information from the database 200 (step S101), and the similarity calculation unit 1102 collects the history information. Analyze history information and compare similarities between users, for example, compare the history of two users based on the number of content with history at the same time, or compare the history of two content and The definition is made based on the number of contents possessed (step S102).
  • the recommendation level calculation unit 1103 uses the calculated similarity and history information necessary for analysis from all the log information to determine the recommendation level of each content for the user, for example, the content history of the user. Calculation is made based on the similarity with other users or the similarity between the contents existing in the user's history and other contents (step S103), and the storage unit 1104 stores the calculated recommendation degree (step S103). S104).
  • the history information collection unit 1201 of the short-term interest analysis processing unit 1200 collects history information necessary for analysis from the log information for each session (step S105), and the similarity calculation unit 1202 collects the collected history information. By analyzing the information, the similarity between users is defined as described above (step S106).
  • the recommendation level calculation unit 1203 calculates the recommendation level of each content for the user using the calculated similarity and history information necessary for analysis from the log information for each session (step S107).
  • the recommendation result output unit 1300 outputs a recommendation result from the stored recommendation level for each content and the recommendation level for each content calculated by the recommendation level calculation unit 1203 according to the change result of the integration ratio calculation unit 1600. (Step S108).
  • the recommendation result is obtained from the recommendation degree using the similarity calculated in the long-term preference analysis process and the recommendation degree using the similarity calculated in the short-term interest analysis process.
  • the recommendation device 1000 includes a long-term preference analysis processing unit 1110, a short-term interest analysis processing unit 1210, a recommendation result output unit 1310, a recommendation display unit 1400, and a browsing situation.
  • the database 200 is connected to the long-term preference analysis processing unit 1110, and the database 300 is connected to the short-term interest analysis processing unit 1210.
  • the recommendation device includes a history information collection unit 1101, a similarity calculation unit 1102, a storage unit 1115, and a short-term interest analysis process that constitute a long-term preference analysis processing unit 1110.
  • the detailed description is abbreviate
  • the storage unit 1115 stores the similarity calculated by the similarity calculation unit 1102.
  • the recommendation level calculation unit 1215 calculates the recommendation level of each content for the user using the similarity stored in the storage unit 1115 and the history information necessary for analysis from the log information for each session.
  • the recommendation result output unit 1310 determines a recommendation result from the recommendation level for each content calculated by the recommendation level calculation unit 1203 and the recommendation level for each content calculated by the recommendation level calculation unit 1215 according to the change result of the integration ratio calculation unit 1600. Is output.
  • the history information collection unit 1101 of the long-term preference analysis processing unit 1110 collects history information necessary for analysis from all log information from the database 200 (step S201), and the similarity calculation unit 1102 collects the history information.
  • the history information is analyzed to define the similarity between users (step S202), and the storage unit 1115 stores the calculated similarity (step S203).
  • the history information collection unit 1201 of the short-term interest analysis processing unit 1210 collects history information necessary for analysis from the database 300 in real time from the log information for each session (step S204), and the similarity calculation unit 1202 collects the history information.
  • the history information is analyzed to define the similarity between users (step S205).
  • the recommendation level calculation unit 1203 calculates the recommendation level of each content for the user using the calculated similarity and history information necessary for analysis from the log information for each session (step S206).
  • the recommendation level calculation unit 1215 calculates the recommendation level of each content for the user using the similarity calculated by the similarity calculation unit 1102 and the history information necessary for analysis from the log information for each session. (Step S207).
  • the recommendation result output unit 1310 uses the recommendation level for each content calculated by the recommendation level calculation unit 1203 and the recommendation level for each content calculated by the recommendation level calculation unit 1215 according to the change result of the integration ratio calculation unit 1600.
  • a recommendation result is output (step S209).
  • the short-term interest analysis processing unit calculates the recommendation degree from the similarity calculated by the long-term preference analysis processing and the similarity calculated by the short-term interest analysis processing.
  • a long-term preference analysis processing unit provides a recommendation result based on the user's long-term preference
  • a short-term interest analysis processing unit provides a short-term preference in the situation.
  • the recommendation apparatus 1000 includes a long-term preference analysis processing unit 1120, a short-term interest analysis processing unit 1220, a recommendation result output unit 1320, a recommendation display unit 1400, and a browsing situation.
  • the database 200 is connected to the long-term preference analysis processing unit 1120
  • the database 300 is connected to the short-term interest analysis processing unit 1220.
  • the recommendation device includes a history information collection unit 1101, a similarity calculation unit 1102, a recommendation degree calculation unit 1103, and a storage unit 1126 that form a long-term preference analysis processing unit 1120.
  • the detailed description is abbreviate
  • the storage unit 1126 stores the similarity defined by the similarity calculation unit 1102 and the recommendation level calculated by the recommendation level calculation unit 1103.
  • the recommendation level calculation unit 1226 calculates the recommendation level of each content for the user using the similarity stored in the storage unit 1126 and the calculated recommendation level and history information necessary for analysis from the log information for each session. To do.
  • the recommendation result output unit 1320 determines a recommendation result from the recommendation level for each content calculated by the recommendation level calculation unit 1103 and the recommendation level for each content calculated by the recommendation level calculation unit 1226 according to the change result of the integration ratio calculation unit 1600. Is output.
  • the history information collection unit 1101 of the long-term preference analysis processing unit 1120 collects history information necessary for analysis from all the log information (step S301), and the similarity calculation unit 1102 collects the history information.
  • the history information is analyzed to define the similarity between users (step S302).
  • the recommendation level calculation unit 1103 calculates the recommendation level of each content for the user using the calculated similarity and history information necessary for analysis from all log information (step S303), and the storage unit 1126
  • the similarity defined in the similarity calculation unit 1102 and the recommendation level calculated in the recommendation level calculation unit 1103 are stored (step S304).
  • the history information collection unit 1201 of the short-term interest analysis processing unit 1220 collects history information necessary for analysis from the log information for each session from the database 300 in real time (step S305).
  • the recommendation level calculation unit 1226 calculates the recommendation level of each content for the user using the similarity and recommendation level stored in the storage unit 1126 and the history information necessary for analysis from the log information for each session ( Step S307).
  • the recommendation result output unit 1320 uses the recommendation level for each content calculated by the recommendation level calculation unit 1103 and the recommendation level for each content calculated by the recommendation level calculation unit 1226 according to the change result of the integration ratio calculation unit 1600. A recommendation result is output (step S308).
  • the recommendation degree using the similarity calculated in the long-term preference analysis process and the short-term interest analysis process use the similarity calculated in the long-term preference analysis process.
  • the long-term preference analysis processing unit provides the recommendation result based on the long-term preference of the user
  • the short-term interest analysis processing unit provides the short-term interest in the situation.
  • the recommendation system includes a long-term preference analysis processing device 400, a short-term interest analysis processing device 500, a recommendation result output device 600, a recommendation display device 700, and browsing status feedback.
  • the long-term preference analysis processing device 400 is connected to the database 200
  • the short-term interest analysis processing device 500 is connected to the database 300.
  • the recommendation system divides the recommendation processing into a long-term preference analysis processing device and a short-term interest analysis processing device, and performs long-term preference analysis processing with the long-term preference analysis processing device.
  • a short-term preference analysis process is performed by the automatic interest analysis processing device, and the result of the long-term preference analysis processing device 400 and the result of the short-term interest analysis processing device 500 are integrated by the recommendation result output device 600 and output.
  • a long-term preference analysis processing device 400 in addition to a cloud server with high calculation processing capability
  • a short-term interest analysis processing device 500 in addition to a cloud server with high calculation processing capability like the long-term preference analysis processing device 400
  • implementation with a portable terminal or STB (Set Top Box) having relatively low arithmetic processing capability can be exemplified, the present invention is not limited to this.
  • the recommendation system includes a history information collection unit 401, a similarity calculation unit 402, a recommendation degree calculation unit 403, and a storage unit 404 that constitute the long-term preference analysis processing device 400.
  • the history information collection unit 401 collects history information necessary for analysis from all the log information from the database 200.
  • the similarity calculation unit 402 analyzes the collected history information and defines the similarity between users.
  • the recommendation level calculation unit 403 calculates the recommendation level of each content for the user using the calculated similarity and history information necessary for analysis collected from all log information.
  • the storage unit 404 stores the calculated recommendation level.
  • the history information collection unit 501 collects history information necessary for analysis from the log information for each session in real time in the database 300.
  • the similarity calculation unit 502 analyzes the collected history information and defines the similarity between users.
  • the recommendation level calculation unit 503 calculates the recommendation level of each content for the user using the calculated similarity and history information necessary for analysis from the log information for each session.
  • the recommendation result output device 600 outputs a recommendation result from the stored recommendation level for each content and the recommendation level for each content calculated by the recommendation level calculation unit 503 according to the change result of the integrated ratio calculation device 900.
  • the function of the integrated ratio calculation apparatus 900 is to change the weighted sum, and is processed by an index / logistic function or the like based on, for example, the number of sessions viewed or the session time.
  • the history information collection unit 401 of the long-term preference analysis processing device 400 collects history information necessary for analysis from all log information from the database 200 (step S401), and the similarity calculation unit 402 collects the history information. Analyze history information and compare similarities between users, for example, compare the history of two users based on the number of content with history at the same time, or compare the history of two content and The definition is made based on the number of contents possessed (step S402).
  • the recommendation level calculation unit 403 uses the calculated similarity and history information necessary for analysis from all log information, and has a recommendation level of each content for the user, for example, a history of the content of the user. Calculation is performed based on the similarity with other users or the similarity between the content existing in the user's history and other content (step S403), and the storage unit 404 stores the calculated recommendation degree (step S403). S404).
  • the history information collection unit 501 of the short-term interest analysis processing apparatus 500 collects history information necessary for analysis from the log information for each session (step S405), and the similarity calculation unit 502 collects the collected history information. By analyzing the information, the similarity between the users is defined as described above (step S406).
  • the recommendation level calculation unit 503 calculates the recommendation level of each content for the user using the calculated similarity and history information necessary for analysis from the log information for each session (step S407).
  • the recommendation result output device 600 outputs a recommendation result from the stored recommendation level for each content and the recommendation level for each content calculated by the recommendation level calculation unit 503 according to the change result of the integration ratio calculation device 900. (Step S408).
  • the degree of recommendation using the similarity calculated in the long-term preference analysis processing device and the similarity calculated by the short-term interest analysis processing device in the long-term preference analysis processing device is output.
  • the long-term preference analysis processing device provides recommendation results based on the user's long-term preference
  • the short-term interest analysis processing device provides a short-term in that situation.
  • the recommendation system includes a long-term preference analysis processing device 410, a short-term interest analysis processing device 510, a recommendation result output device 610, a recommendation display device 700, and browsing status feedback.
  • the long-term preference analysis processing device 410 is connected to the database 200
  • the short-term interest analysis processing device 510 is connected to the database 300.
  • the recommendation system includes a history information collection unit 401, a similarity calculation unit 402, a storage unit 415, and a short-term interest analysis process that constitute the long-term preference analysis processing device 410.
  • the browsing status feedback device 800 and the integrated ratio calculation device 900 are configured.
  • symbol similar to 4th Embodiment, since it has the same function the detailed description is abbreviate
  • the storage unit 415 stores the similarity calculated by the similarity calculation unit 402.
  • the recommendation level calculation unit 515 calculates the recommendation level of each content for the user using the similarity stored in the storage unit 415 and the history information necessary for analysis from the log information for each session.
  • the recommendation result output device 610 determines a recommendation result from the recommendation level for each content calculated by the recommendation level calculation unit 503 and the recommendation level for each content calculated by the recommendation level calculation unit 515 according to the change result of the integration ratio calculation device 900. Is output.
  • the history information collection unit 401 of the long-term preference analysis processing apparatus 410 collects history information necessary for analysis from all log information from the database 200 (step S501), and the similarity calculation unit 402 collects the history information.
  • the history information is analyzed to define the similarity between users (step S502), and the storage unit 415 stores the calculated similarity (step S503).
  • the history information collection unit 501 of the short-term interest analysis processing apparatus 510 collects history information necessary for analysis from the log information for each session in real time from the database 300 (step S504), and the similarity calculation unit 502 collects the history information.
  • the history information is analyzed to define similarity between users (step S505).
  • the recommendation level calculation unit 503 calculates the recommendation level of each content for the user using the calculated similarity and history information necessary for analysis from the log information for each session (step S506).
  • the recommendation level calculation unit 515 calculates the recommendation level of each content for the user using the similarity calculated by the similarity calculation unit 402 and the history information necessary for analysis from the log information for each session. (Step S507).
  • the recommendation result output device 610 uses the recommendation level for each content calculated by the recommendation level calculation unit 503 and the recommendation level for each content calculated by the recommendation level calculation unit 515 according to the change result of the integrated ratio calculation device 900.
  • a recommendation result is output (step S509).
  • the degree of recommendation using the similarity calculated in the long-term preference analysis processing device and the similarity calculated by the short-term interest analysis processing device in the long-term preference analysis processing device is output.
  • the long-term preference analysis processing device provides recommendation results based on the user's long-term preference
  • the short-term interest analysis processing device provides a short-term in that situation.
  • the recommendation system includes a long-term preference analysis processing device 420, a short-term interest analysis processing device 520, a recommendation result output device 620, a recommendation display device 700, and browsing status feedback.
  • the long-term preference analysis processing device 420 is connected to the database 200
  • the short-term interest analysis processing device 520 is connected to the database 300.
  • the recommendation system includes a history information collection unit 401, a similarity calculation unit 402, a recommendation degree calculation unit 403, and a storage unit 426 that constitute the long-term preference analysis processing device 420.
  • an integrated ratio calculation device 900 is an integrated ratio calculation device 900.
  • the storage unit 426 stores the similarity defined by the similarity calculation unit 402 and the recommendation level calculated by the recommendation level calculation unit 403.
  • the recommendation level calculation unit 526 calculates the recommendation level of each content for the user using the similarity stored in the storage unit 426 and the calculated recommendation level and history information necessary for analysis from the log information for each session. To do.
  • the recommendation result output device 620 determines a recommendation result from the recommendation level for each content calculated by the recommendation level calculation unit 403 and the recommendation level for each content calculated by the recommendation level calculation unit 526 according to the change result of the integration ratio calculation device 900. Is output.
  • the history information collection unit 401 of the long-term preference analysis processing device 420 collects history information necessary for analysis from all log information from the database 200 (step S601), and the similarity calculation unit 402 collects the history information.
  • the history information is analyzed to define the similarity between users (step S602).
  • the recommendation level calculation unit 403 calculates the recommendation level of each content for the user using the calculated similarity and history information necessary for analysis from all the log information (step S603), and the storage unit 426
  • the similarity defined in the similarity calculation unit 402 and the recommendation level calculated in the recommendation level calculation unit 403 are stored (step S604).
  • the history information collection unit 501 of the short-term interest analysis processing device 520 collects history information necessary for analysis from the database 300 in real time from the log information for each session (step S605).
  • the recommendation level calculation unit 526 calculates the recommendation level of each content for the user using the similarity and recommendation level stored in the storage unit 426 and the history information necessary for analysis from the log information for each session ( Step S607).
  • the recommendation result output device 620 determines from the recommendation level for each content calculated by the recommendation level calculation unit 403 and the recommendation level for each content calculated by the recommendation level calculation unit 526 according to the change result of the integrated ratio calculation device 900.
  • a recommendation result is output (step S608).
  • the degree of recommendation using the similarity calculated in the long-term preference analysis processing device and the similarity calculated by the short-term interest analysis processing device in the long-term preference analysis processing device is output.
  • the long-term preference analysis processing device provides recommendation results based on the user's long-term preference
  • the short-term interest analysis processing device provides a short-term in that situation.
  • the processing of the recommendation device or the recommendation system is recorded on a computer-readable recording medium, and the program recorded on the recording medium is recommended, the long-term preference analysis processing device, the short-term interest analysis processing device, or the recommendation result output device.
  • the recommendation device or the recommendation system of the present invention can be realized by reading and executing the above.
  • the computer system here includes an OS and hardware such as peripheral devices.
  • the “computer system” includes a homepage providing environment (or display environment) if a WWW (World Wide Web) system is used.
  • the program may be transmitted from a computer system storing the program in a storage device or the like to another computer system via a transmission medium or by a transmission wave in the transmission medium.
  • the “transmission medium” for transmitting the program refers to a medium having a function of transmitting information, such as a network (communication network) such as the Internet or a communication line (communication line) such as a telephone line.
  • the program may be for realizing a part of the above-described functions. Furthermore, what can implement
  • history information to be used is centrally managed in a storage medium such as a database, and it is assumed that these are acquired using a history information collection unit.
  • the present invention is not necessarily limited to this.
  • the method for providing recommendation results according to the present invention assumes that a list describing the degree of recommendation of each content for the user is provided to an external recommendation presentation method. It is not limited to.
  • each process of the recommendation according to the present invention may be performed by one server, or by an implementation method such as implementation by cooperation of a plurality of servers specialized for each function or cooperation of a plurality of servers using a load distribution function. .
  • recommendation device 1100 long-term preference analysis processing unit 1110: long-term preference analysis processing unit 1120: long-term preference analysis processing unit 1101: history information collection unit 1102: similarity calculation unit 1103: recommendation degree calculation unit 1104: storage unit 1115: Storage unit 1126: Storage unit 1200: Short-term interest analysis processing unit 1210: Short-term interest analysis processing unit 1220: Short-term interest analysis processing unit 1201: History information collection unit 1202: Similarity calculation unit 1203: Recommendation degree calculation unit 1215: Recommendation degree calculation unit 1226: Recommendation degree calculation unit 1300: Recommendation result output unit 1310: Recommendation result output unit 1320: Recommendation result output unit 1400: Recommendation display unit 1500: Browsing status feedback unit 1600: Integration ratio calculation unit 200: Database 300: Database 4 0: Long-term preference analysis processing device 410: Long-term preference analysis processing device 420: Long-term preference analysis processing device 401: History information collection unit 402: Similarity calculation

Landscapes

  • Business, Economics & Management (AREA)
  • Engineering & Computer Science (AREA)
  • Finance (AREA)
  • Accounting & Taxation (AREA)
  • Strategic Management (AREA)
  • Development Economics (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Marketing (AREA)
  • Economics (AREA)
  • General Business, Economics & Management (AREA)
  • Databases & Information Systems (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Data Mining & Analysis (AREA)
  • Game Theory and Decision Science (AREA)
  • General Engineering & Computer Science (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention addresses the problem of responding accurately to an item in which a user has had a short-term interest, and achieving a high-precision recommendation. To do so, a long-term preference analysis processing means analyzes an entirety of history information, and generates information needed for the recommendation. A short-term interest analysis processing means analyzes session-specific history information, and generates information needed for the recommendation. A recommendation result output means integrates the output of the long-term preference analysis processing means and/or the output of the short-term interest analysis processing means, and outputs the recommendation result. On the basis of a change in the user's interest, an integration proportion computation means modifies the recommendation result of the recommendation result output means.

Description

レコメンド装置、レコメンドシステム、レコメンド方法およびプログラムRECOMMENDATION DEVICE, RECOMMENDATION SYSTEM, RECOMMENDATION METHOD, AND PROGRAM
　本発明は、長期的嗜好分析処理と短期的興味分析処理とを行い、両処理のレコメンド結果を統合して、最終的なレコメンド結果を出力するレコメンド装置、レコメンドシステム、レコメンド方法およびプログラムに関する。 The present invention relates to a recommendation device, a recommendation system, a recommendation method, and a program that perform a long-term preference analysis process and a short-term interest analysis process, integrate the recommendation results of both processes, and output a final recommendation result.
　近年、ユーザの履歴情報などを基にユーザの嗜好を推測し、ユーザの嗜好に適合すると推測されるコンテンツなどを自動的にレコメンドする「レコメンドサービス」が盛んに行われている。このようなレコメンドサービスにおいて利用されている代表的な技術が「協調フィルタリング」と呼ばれているものである。この協調フィルタリングは、購買履歴や閲覧履歴等のユーザの履歴情報を、各ユーザ間、あるいは各コンテンツ間で比較することでユーザ間の類似性を定義し、その類似性を元にユーザが興味を持つと予測されるコンテンツをレコメンド結果として出力する手法である。この協調フィルタリングについては、自動化や高速化等の改良技術が知られている（例えば、非特許文献１、非特許文献２、非特許文献３、非特許文献４参照。）。 In recent years, “recommendation services” have been actively performed in which the user's preference is estimated based on the user's history information and the like, and content that is presumed to match the user's preference is automatically recommended. A typical technique used in such a recommendation service is called “collaborative filtering”. This collaborative filtering defines the similarity between users by comparing user history information such as purchase history and browsing history between each user or between each content, and the user is interested based on the similarity. This is a technique for outputting content predicted to have as a recommendation result. For this collaborative filtering, improved techniques such as automation and high speed are known (see, for example, Non-Patent Document 1, Non-Patent Document 2, Non-Patent Document 3, and Non-Patent Document 4).
　具体的には、非特許文献１においては、自動化された協調フィルタリングが提案されており、非特許文献２等では、確率モデルに基づく協調フィルタリングが提案され、一方、非特許文献３では、非特許文献１の協調フィルタリングの高速化が試みられた。また、非特許文献４は、ユーザの履歴情報において、嗜好の減衰モデルを仮定することで、最新の履歴情報を重視可能な協調フィルタリング手法を提案している。 Specifically, in Non-Patent Document 1, automated collaborative filtering is proposed. In Non-Patent Document 2, etc., collaborative filtering based on a probability model is proposed, while in Non-Patent Document 3, non-patent An attempt was made to speed up the collaborative filtering of Document 1. Further, Non-Patent Document 4 proposes a collaborative filtering technique capable of placing importance on the latest history information by assuming a preference decay model in the user history information.
　しかしながら、非特許文献１や非特許文献２に記載の技術では、蓄積され続けた履歴情報すべてを対象に解析してレコメンド結果を生成するため、ユーザにおいて必要以上に履歴が蓄積されると、レコメンド処理によって算出されるコンテンツが毎回固定的となり、新しいコンテンツの発見ができなくなるという問題点があった。 However, in the techniques described in Non-Patent Document 1 and Non-Patent Document 2, since all the history information that has been accumulated is analyzed and a recommendation result is generated, if the history is accumulated more than necessary by the user, the recommendation The content calculated by the processing is fixed each time, and there is a problem that new content cannot be found.
　また、非特許文献３に記載の技術では、代表ユーザと被推薦対象ユーザとの間でのみ類似性を定義するため、非特許文献１に記載の技術と比較して推薦要求時に解析する計算量が小さくなり、最新の履歴情報を元にレコメンドの処理を実行することを容易にするが、非特許文献３では、その推薦処理の高速化のみに着目しており、最新の履歴情報を考慮してレコメンドの提示内容を変更する等の処理はされておらず、前述同様にユーザにおいて必要以上に履歴が蓄積されると、レコメンドによって提示されるコンテンツが毎回固定的となる問題が依然として残っている。 Further, in the technique described in Non-Patent Document 3, since the similarity is defined only between the representative user and the user to be recommended, the amount of calculation to be analyzed at the time of recommendation compared with the technique described in Non-Patent Document 1 However, Non-Patent Document 3 focuses only on speeding up the recommendation process and considers the latest history information. However, if the history is accumulated more than necessary by the user as described above, there is still a problem that the content presented by the recommendation is fixed every time. .
　さらに、非特許文献４では、最新の履歴情報を重視することで、ユーザの嗜好の移り変わりを許容できるようにしているが、当該ユーザの瞬間的な興味の盛り上がりを考慮するものではなく、より最新の履歴情報を頼りに嗜好を分析しようとするものである。そのため、ユーザの長期的な嗜好から外れた、その場限りの興味に対するレコメンドに対しての追従性が低いという問題がある。 Furthermore, in Non-Patent Document 4, emphasis is placed on the latest history information so as to allow a change in the user's preference, but it does not take into account the momentary excitement of the user, and is more up-to-date. It is intended to analyze the preference based on the history information. Therefore, there is a problem that followability to a recommendation with respect to a one-time interest that is out of the long-term preference of the user is low.
　つまり、上述の従来技術を組み合わせたとしても、履歴情報が十分にある場合には、協調フィルタリングにおける処理結果が毎回、同じものになってしまう可能性が高く、仮に、履歴情報が数個変わったとしても、その大意が変化しないため、ある瞬間に興味を感じた分野に応じてレコメンド結果を動的に変更したいというニーズに応えることが出来ない。 That is, even if the above-described conventional techniques are combined, if there is enough history information, the processing result in collaborative filtering is likely to be the same every time, and there are several changes in the history information. However, since the intention does not change, it is not possible to meet the need to dynamically change the recommendation result according to the field that interestd him at a certain moment.
　そこで、本発明は、上述の課題に鑑みてなされたものであり、ユーザが短期的に興味を持ったものに対して、的確に対応し、精度の高いレコメンドを実現するレコメンド装置、レコメンドシステム、レコメンド方法およびプログラムを提供することを目的とする。 Therefore, the present invention has been made in view of the above-described problems, and a recommendation device, a recommendation system, which achieves a highly accurate recommendation that accurately corresponds to what the user is interested in in the short term, It is an object to provide a recommendation method and program.
　本発明は、上記の課題を解決するために、以下の事項を提案している。なお、理解を容易にするために、本発明の実施形態に対応する符号を付して説明するが、これに限定されるものではない。 The present invention proposes the following items in order to solve the above problems. In addition, in order to make an understanding easy, although the code | symbol corresponding to embodiment of this invention is attached | subjected and demonstrated, it is not limited to this.
　（１）本発明は、購買履歴や閲覧履歴等のユーザの履歴情報を元にユーザ間の類似性を定義し、該定義した類似性に基づいてユーザが興味を持つと予測されるコンテンツをレコメンド結果として出力する協調フィルタリング方法を用いるレコメンド装置であって、すべての前記履歴情報を解析してレコメンドに必要な情報を生成する長期的嗜好分析処理手段（例えば、図１の長期的嗜好分析処理部１１００に相当）と、セッションごとの履歴情報を解析してレコメンドに必要な情報を生成する短期的興味分析処理手段（例えば、図１の短期的興味分析処理部１２００に相当）と、前記長期的嗜好分析処理手段の出力または／および前記短期的興味分析処理手段の出力を統合してレコメンド結果を出力するレコメンド結果出力手段（例えば、図１のレコメンド結果出力部１３００に相当）と、前記レコメンド結果出力手段のレコメンド結果をユーザの興味の遷移に基づいて変更する統合割合算出手段（例えば、図１の統合割合算出部１６００に相当）と、を備えたことを特徴とするレコメンド装置を提案している。 (1) The present invention defines similarity between users based on user history information such as purchase history and browsing history, and recommends content that is predicted to be of interest to the user based on the defined similarity. A recommendation device that uses the collaborative filtering method to output as a result, which analyzes all the history information and generates information necessary for the recommendation (for example, the long-term preference analysis processing unit in FIG. 1) 1100), short-term interest analysis processing means (for example, equivalent to the short-term interest analysis processing unit 1200 in FIG. 1) for analyzing history information for each session and generating information necessary for recommendation, and the long-term Recommendation result output means (for example, output of recommendation results by integrating the output of the preference analysis processing means and / or the output of the short-term interest analysis processing means (for example, 1 corresponding to the recommendation result output unit 1300), and an integration ratio calculation unit that changes the recommendation result of the recommendation result output unit based on the transition of the user's interest (for example, corresponding to the integration ratio calculation unit 1600 in FIG. 1); The recommendation apparatus characterized by having provided these.
　この発明によれば、長期的嗜好分析処理手段は、すべての履歴情報を解析してレコメンドに必要な情報を生成する。短期的興味分析処理手段は、セッションごとの履歴情報を解析してレコメンドに必要な情報を生成する。レコメンド結果出力手段は、長期的嗜好分析処理手段の出力または／および短期的興味分析処理手段の出力を統合してレコメンド結果を出力する。統合割合算出手段は、レコメンド結果出力手段のレコメンド結果をユーザの興味の遷移に基づいて変更する。したがって、長期的嗜好分析処理手段によって、ユーザの長期的な嗜好に基づくレコメンド結果を提供し、かつ、短期的興味分析処理手段によってその状況においての短期的な興味に対するレコメンド結果を提供することで、興味の遷移を考慮した適応的なレコメンド結果を提供することが可能となる。 According to this invention, the long-term preference analysis processing means analyzes all history information and generates information necessary for the recommendation. The short-term interest analysis processing means analyzes history information for each session and generates information necessary for recommendation. The recommendation result output means integrates the output of the long-term preference analysis processing means and / or the output of the short-term interest analysis processing means, and outputs a recommendation result. The integration ratio calculation unit changes the recommendation result of the recommendation result output unit based on the transition of the user's interest. Therefore, by providing a recommendation result based on the long-term preference of the user by the long-term preference analysis processing means, and by providing a recommendation result for the short-term interest in the situation by the short-term interest analysis processing means, It is possible to provide an adaptive recommendation result considering the transition of interest.
　（２）本発明は、（１）のレコメンド装置について、前記長期的嗜好分析処理手段が、すべてのログ情報の中から解析に必要な履歴情報を収集する第１の履歴情報収集手段（例えば、図２の履歴情報収集部１１０１に相当）と、該収集した履歴情報を解析して、ユーザ間の類似性を定義する第１の類似性算出手段（例えば、図２の類似性算出部１１０２に相当）と、該算出された類似性と該すべてのログ情報の中から収集された解析に必要な履歴情報とを用いて、ユーザに対する各コンテンツの推薦度を算出する第１の推薦度算出手段（例えば、図２の推薦度算出部１１０３に相当）と、該算出した推薦度を記憶する記憶手段（例えば、図２の記憶部１１０４に相当）と、を備え、前記短期的興味分析処理手段が、セッションごとのログ情報の中から解析に必要な履歴情報をリアルタイムに収集する第２の履歴情報収集手段（例えば、図２の履歴情報収集部１２０１に相当）と、該収集した履歴情報を解析して、ユーザ間の類似性を定義する第２の類似性算出手段（例えば、図２の類似性算出部１２０２に相当）と、該算出された類似性と該セッションごとのログ情報の中から解析に必要な履歴情報とを用いて、ユーザに対する各コンテンツの推薦度を算出する第２の推薦度算出手段（例えば、図２の推薦度算出部１２０３に相当）と、を備え、前記レコメンド結果出力手段が、前記統合割合算出手段の変更結果に応じて、前記記憶されたコンテンツごとの推薦度と前記第２の推薦度算出手段により算出したコンテンツごとの推薦度とからレコメンド結果を出力することを特徴とするレコメンド装置を提案している。 (2) In the recommendation device according to (1), the long-term preference analysis processing means collects history information necessary for analysis from all log information (for example, first history information collection means (for example, 2) and first similarity calculation means for analyzing the collected history information and defining the similarity between users (for example, the similarity calculation unit 1102 in FIG. 2). First recommendation degree calculating means for calculating the recommendation degree of each content for the user using the calculated similarity and the history information necessary for the analysis collected from all the log information (For example, equivalent to the recommendation degree calculation unit 1103 in FIG. 2) and storage means for storing the calculated recommendation degree (for example, equivalent to the storage part 1104 in FIG. 2), the short-term interest analysis processing means Log information for each session The second history information collection means (for example, equivalent to the history information collection unit 1201 in FIG. 2) that collects the history information necessary for analysis from among the users, and analyzes the collected history information, Second similarity calculation means for defining similarity (for example, equivalent to the similarity calculation unit 1202 in FIG. 2), history information necessary for analysis from the calculated similarity and log information for each session And a second recommendation level calculation unit (for example, equivalent to the recommendation level calculation unit 1203 in FIG. 2) that calculates the recommendation level of each content for the user, and the recommendation result output unit includes the integration According to the change result of the ratio calculation means, a recommendation result is output from the stored recommendation degree for each content and the recommendation degree for each content calculated by the second recommendation degree calculation means, It has proposed that recommendation apparatus.
　この発明によれば、長期的嗜好分析処理手段の第１の履歴情報収集手段は、すべてのログ情報の中から解析に必要な履歴情報を収集する。第１の類似性算出手段は、収集した履歴情報を解析して、ユーザ間の類似性を定義する。第１の推薦度算出手段は、算出された類似性とすべてのログ情報の中から収集された解析に必要な履歴情報とを用いて、ユーザに対する各コンテンツの推薦度を算出する。記憶手段は、算出した推薦度を記憶する。短期的興味分析処理手段の第２の履歴情報収集手段は、セッションごとのログ情報の中から解析に必要な履歴情報をリアルタイムに収集する。第２の類似性算出手段は、収集した履歴情報を解析して、ユーザ間の類似性を定義する。第２の推薦度算出手段は、算出された類似性とセッションごとのログ情報の中から解析に必要な履歴情報とを用いて、ユーザに対する各コンテンツの推薦度を算出する。レコメンド結果出力手段は、統合割合算出手段の変更結果に応じて、記憶されたコンテンツごとの推薦度と第２の推薦度算出手段により算出したコンテンツごとの推薦度とからレコメンド結果を出力する。したがって、長期的嗜好分析処理手段によって、ユーザの長期的な嗜好に基づくレコメンド結果を提供し、かつ、短期的興味分析処理手段によってその状況においての短期的な興味に対するレコメンド結果を提供することで、興味の遷移を考慮した適応的なレコメンド結果を提供することが可能となる。 According to this invention, the first history information collection means of the long-term preference analysis processing means collects history information necessary for analysis from all log information. The first similarity calculation unit analyzes the collected history information and defines the similarity between users. The first recommendation degree calculating means calculates the recommendation degree of each content for the user using the calculated similarity and the history information necessary for the analysis collected from all the log information. The storage means stores the calculated recommendation level. The second history information collection means of the short-term interest analysis processing means collects history information necessary for analysis from the log information for each session in real time. The second similarity calculation means analyzes the collected history information and defines the similarity between users. The second recommendation degree calculation means calculates the recommendation degree of each content for the user using the calculated similarity and history information necessary for analysis from the log information for each session. The recommendation result output unit outputs a recommendation result from the stored recommendation level for each content and the recommendation level for each content calculated by the second recommendation level calculation unit according to the change result of the integration ratio calculation unit. Therefore, by providing a recommendation result based on the long-term preference of the user by the long-term preference analysis processing means, and by providing a recommendation result for the short-term interest in the situation by the short-term interest analysis processing means, It is possible to provide an adaptive recommendation result considering the transition of interest.
　（３）本発明は、（１）のレコメンド装置について、前記長期的嗜好分析処理手段が、すべてのログ情報の中から解析に必要な履歴情報を収集する第１の履歴情報収集手段（例えば、図５の履歴情報収集部１１０１に相当）と、該収集した履歴情報を解析して、ユーザ間の類似性を定義する第１の類似性算出手段（例えば、図５の類似性算出部１１０２に相当）と、該算出した類似性を記憶する記憶手段（例えば、図５の記憶部１１０４に相当）と、を備え、前記短期的興味分析処理手段が、セッションごとのログ情報の中から解析に必要な履歴情報をリアルタイムに収集する第２の履歴情報収集手段（例えば、図５の履歴情報収集部１２０１に相当）と、該収集した履歴情報を解析して、ユーザ間の類似性を定義する第２の類似性算出手段（例えば、図５の類似性算出部１２０２に相当）と、該セッションごとのログ情報の中から解析に必要な履歴情報と前記記憶された類似性とを用いて、ユーザに対する各コンテンツの推薦度を算出する第１の推薦度算出手段（例えば、図５の推薦度算出部１２０３に相当）と、前記第２の類似性算出手段により算出された類似性と前記セッションごとのログ情報の中から解析に必要な履歴情報とを用いて、ユーザに対する各コンテンツの推薦度を算出する第２の推薦度算出手段（例えば、図５の推薦度算出部１２１５に相当）と、を備え、前記レコメンド結果出力手段が、前記統合割合算出手段の変更結果に応じて、前記第１の推薦度算出手段により算出したコンテンツごとの推薦度と前記第２の推薦度算出手段により算出したコンテンツごとの推薦度とからレコメンド結果を出力することを特徴とするレコメンド装置を提案している。 (3) According to the present invention, in the recommendation device of (1), the long-term preference analysis processing means collects history information necessary for analysis from all log information (for example, first history information collection means (for example, 5) and a first similarity calculation unit that analyzes the collected history information and defines the similarity between users (for example, the similarity calculation unit 1102 in FIG. 5). And a storage means (for example, equivalent to the storage unit 1104 in FIG. 5) for storing the calculated similarity, and the short-term interest analysis processing means analyzes the log information for each session. Second history information collection means (for example, equivalent to the history information collection unit 1201 in FIG. 5) that collects necessary history information in real time, and analyzes the collected history information to define similarity between users Second similarity calculator (For example, equivalent to the similarity calculation unit 1202 in FIG. 5) and the history information necessary for analysis from the log information for each session and the stored similarity, the degree of recommendation of each content to the user From the first recommendation degree calculation means (for example, equivalent to the recommendation degree calculation unit 1203 in FIG. 5), the similarity calculated by the second similarity calculation means, and the log information for each session Second recommendation level calculation means (for example, equivalent to the recommendation level calculation unit 1215 in FIG. 5) that calculates the recommendation level of each content for the user using history information necessary for analysis, and the recommendation result The output unit calculates the recommendation level for each content calculated by the first recommendation level calculation unit and the content calculated by the second recommendation level calculation unit according to the change result of the integration ratio calculation unit. It proposes a recommendation apparatus and outputting a recommendation result from a recommendation degree with.
　この発明によれば、長期的嗜好分析処理手段の第１の履歴情報収集手段は、すべてのログ情報の中から解析に必要な履歴情報を収集する。第１の類似性算出手段は、収集した履歴情報を解析して、ユーザ間の類似性を定義する。記憶手段は、算出した類似性を記憶する。短期的興味分析処理手段の第２の履歴情報収集手段は、セッションごとのログ情報の中から解析に必要な履歴情報をリアルタイムに収集する。第２の類似性算出手段は、収集した履歴情報を解析して、ユーザ間の類似性を定義する。第１の推薦度算出手段は、セッションごとのログ情報の中から解析に必要な履歴情報と記憶された類似性とを用いて、ユーザに対する各コンテンツの推薦度を算出する。第２の推薦度算出手段は、第２の類似性算出手段により算出された類似性とセッションごとのログ情報の中から解析に必要な履歴情報とを用いて、ユーザに対する各コンテンツの推薦度を算出する。レコメンド結果出力手段は、統合割合算出手段の変更結果に応じて、第１の推薦度算出手段により算出したコンテンツごとの推薦度と第２の推薦度算出手段により算出したコンテンツごとの推薦度とからレコメンド結果を出力する。したがって、長期的嗜好分析処理手段によって、ユーザの長期的な嗜好に基づくレコメンド結果を提供し、かつ、短期的興味分析処理手段によってその状況においての短期的な興味に対するレコメンド結果を提供することで、興味の遷移を考慮した適応的なレコメンド結果を提供することが可能となる。 According to this invention, the first history information collection means of the long-term preference analysis processing means collects history information necessary for analysis from all log information. The first similarity calculation unit analyzes the collected history information and defines the similarity between users. The storage means stores the calculated similarity. The second history information collection means of the short-term interest analysis processing means collects history information necessary for analysis from the log information for each session in real time. The second similarity calculation means analyzes the collected history information and defines the similarity between users. The first recommendation level calculation means calculates the recommendation level of each content for the user by using the history information necessary for the analysis and the stored similarity from the log information for each session. The second recommendation level calculation means uses the similarity calculated by the second similarity calculation means and the history information necessary for analysis from the log information for each session to determine the recommendation level of each content to the user. calculate. The recommendation result output means is based on the recommendation degree for each content calculated by the first recommendation degree calculation means and the recommendation degree for each content calculated by the second recommendation degree calculation means according to the change result of the integration ratio calculation means. Output recommendation results. Therefore, by providing a recommendation result based on the long-term preference of the user by the long-term preference analysis processing means, and by providing a recommendation result for the short-term interest in the situation by the short-term interest analysis processing means, It is possible to provide an adaptive recommendation result considering the transition of interest.
　（４）本発明は、（１）のレコメンド装置について、前記長期的嗜好分析処理手段が、すべてのログ情報の中から解析に必要な履歴情報を収集する第１の履歴情報収集手段（例えば、図８の履歴情報収集部１１０１に相当）と、該収集した履歴情報を解析して、ユーザ間の類似性を定義する第１の類似性算出手段（例えば、図８の類似性算出部１１０２に相当）と、該算出された類似性と該セッションごとのログ情報の中から解析に必要な履歴情報とを用いて、ユーザに対する各コンテンツの推薦度を算出する第１の推薦度算出手段（例えば、図８の推薦度算出部１１０３に相当）と、該定義した類似性および算出した推薦度を記憶する記憶手段と、を備え、前記短期的興味分析処理手段が、セッションごとのログ情報の中から解析に必要な履歴情報をリアルタイムに収集する第２の履歴情報収集手段（例えば、図８の履歴情報収集部１２０１に相当）と、前記定義した類似性および算出した推薦度と前記セッションごとのログ情報の中から解析に必要な履歴情報とを用いて、ユーザに対する各コンテンツの推薦度を算出する第２の推薦度算出手段（例えば、図８の推薦度算出部１２２６に相当）と、前記レコメンド結果出力手段が、前記統合割合算出手段の変更結果に応じて、前記第１の推薦度算出手段により算出したコンテンツごとの推薦度と前記第２の推薦度算出手段により算出したコンテンツごとの推薦度とからレコメンド結果を出力することを特徴とするレコメンド装置を提案している。 (4) According to the present invention, for the recommendation device of (1), the long-term preference analysis processing means collects history information necessary for analysis from all log information (for example, first history information collection means (for example, 8) and first similarity calculation means for analyzing the collected history information and defining the similarity between users (for example, the similarity calculation unit 1102 in FIG. 8). First recommendation degree calculation means (for example, a degree of recommendation of each content for the user) using the calculated similarity and history information necessary for analysis from the log information for each session. , Corresponding to the recommendation degree calculation unit 1103 in FIG. 8) and storage means for storing the defined similarity and the calculated recommendation degree. The short-term interest analysis processing means includes log information for each session. Required for analysis Second history information collecting means for collecting history information in real time (for example, corresponding to the history information collecting unit 1201 in FIG. 8), the defined similarity, the calculated recommendation level, and the log information for each session Second recommendation degree calculation means (for example, equivalent to the recommendation degree calculation unit 1226 in FIG. 8) that calculates the recommendation degree of each content for the user using history information necessary for the analysis, and the recommendation result output means A recommendation result from the recommendation level for each content calculated by the first recommendation level calculation unit and the recommendation level for each content calculated by the second recommendation level calculation unit according to the change result of the integration ratio calculation unit. We propose a recommendation device characterized by outputting.
　この発明によれば、長期的嗜好分析処理手段の第１の履歴情報収集手段は、すべてのログ情報の中から解析に必要な履歴情報を収集する。第１の類似性算出手段は、収集した履歴情報を解析して、ユーザ間の類似性を定義する。第１の推薦度算出手段は、算出された類似性とセッションごとのログ情報の中から解析に必要な履歴情報とを用いて、ユーザに対する各コンテンツの推薦度を算出する。記憶手段は、定義した類似性および算出した推薦度を記憶する。短期的興味分析処理手段の第２の履歴情報収集手段は、セッションごとのログ情報の中から解析に必要な履歴情報をリアルタイムに収集する。第２の推薦度算出手段は、定義した類似性および算出した推薦度とセッションごとのログ情報の中から解析に必要な履歴情報とを用いて、ユーザに対する各コンテンツの推薦度を算出する。レコメンド結果出力手段は、統合割合算出手段の変更結果に応じて、第１の推薦度算出手段により算出したコンテンツごとの推薦度と第２の推薦度算出手段により算出したコンテンツごとの推薦度とからレコメンド結果を出力する。したがって、長期的嗜好分析処理手段によって、ユーザの長期的な嗜好に基づくレコメンド結果を提供し、かつ、短期的興味分析処理手段によってその状況においての短期的な興味に対するレコメンド結果を提供することで、興味の遷移を考慮した適応的なレコメンド結果を提供することが可能となる。 According to this invention, the first history information collection means of the long-term preference analysis processing means collects history information necessary for analysis from all log information. The first similarity calculation unit analyzes the collected history information and defines the similarity between users. The first recommendation level calculating means calculates the recommendation level of each content for the user using the calculated similarity and history information necessary for analysis from the log information for each session. The storage means stores the defined similarity and the calculated recommendation level. The second history information collection means of the short-term interest analysis processing means collects history information necessary for analysis from the log information for each session in real time. The second recommendation degree calculation means calculates the recommendation degree of each content for the user using the defined similarity, the calculated recommendation degree, and history information necessary for analysis from the log information for each session. The recommendation result output means is based on the recommendation degree for each content calculated by the first recommendation degree calculation means and the recommendation degree for each content calculated by the second recommendation degree calculation means according to the change result of the integration ratio calculation means. Output recommendation results. Therefore, by providing a recommendation result based on the long-term preference of the user by the long-term preference analysis processing means, and by providing a recommendation result for the short-term interest in the situation by the short-term interest analysis processing means, It is possible to provide an adaptive recommendation result considering the transition of interest.
　（５）本発明は、（１）から（４）のレコメンド装置について、前記セッションごとのログ情報を格納するデータベースと、前記レコメンド結果出力手段から出力されるレコメンドについて、ユーザの閲覧状況を取得する閲覧状況取得手段と、該取得したユーザの閲覧状況を前記データベースと前記統合割合算出手段にフィードバックする閲覧状況フィードバック手段（例えば、図１の閲覧状況フィードバック部１５００に相当）と、を備え、前記統合割合算出手段が閲覧状況フィードバック情報に基づいて、前記レコメンド結果出力手段のレコメンド結果をユーザの興味の遷移に基づいて変更することを特徴とするレコメンド装置を提案している。 (5) The present invention acquires the browsing status of the user with respect to the recommendation device of (1) to (4), with respect to the database storing the log information for each session and the recommendation output from the recommendation result output means. Browsing status acquisition means; and browsing status feedback means for feeding back the acquired browsing status of the user to the database and the integrated ratio calculation means (for example, corresponding to the browsing status feedback unit 1500 in FIG. 1), and the integration The recommendation device is characterized in that the ratio calculation means changes the recommendation result of the recommendation result output means based on the transition of the user's interest based on the browsing status feedback information.
　この発明によれば、データベースは、セッションごとのログ情報を格納する。閲覧状況取得手段は、レコメンド結果出力手段から出力されるレコメンドについて、ユーザの閲覧状況を取得する。閲覧状況フィードバック手段は、取得したユーザの閲覧状況をデータベースと統合割合算出手段にフィードバックする。統合割合算出手段は、閲覧状況フィードバック情報に基づいて、レコメンド結果出力手段のレコメンド結果をユーザの興味の遷移に基づいて変更する。したがって、取得したユーザの閲覧状況をデータベースと統合割合算出手段にフィードバックし、統合割合算出手段が、閲覧状況フィードバック情報に基づいて、レコメンド結果出力手段のレコメンド結果をユーザの興味の遷移に基づいて変更するため、ユーザが短期的に興味をもったコンテンツ等を的確に抽出して、レコメンドを行うことができる。 According to the present invention, the database stores log information for each session. The browsing status acquisition unit acquires the browsing status of the user for the recommendation output from the recommendation result output unit. The browsing status feedback means feeds back the acquired browsing status of the user to the database and the integrated ratio calculation means. The integrated ratio calculation means changes the recommendation result of the recommendation result output means based on the transition of the user's interest based on the browsing status feedback information. Therefore, the obtained browsing status of the user is fed back to the database and the integrated ratio calculating means, and the integrated ratio calculating means changes the recommended result of the recommended result output means based on the transition of the user's interest based on the browsing status feedback information. Therefore, it is possible to make a recommendation by accurately extracting contents and the like that the user is interested in in the short term.
　（６）本発明は、購買履歴や閲覧履歴等のユーザの履歴情報を元にユーザ間の類似性を定義し、該定義した類似性に基づいてユーザが興味を持つと予測されるコンテンツをレコメンド結果として出力する協調フィルタリング方法を用いるレコメンドシステムであって、すべての前記履歴情報を格納する第１のデータベース（例えば、図１０のデータベース２００に相当）と、セッションごとの前記履歴情報を格納する第２のデータベース（例えば、図１０のデータベース３００に相当）と、前記第１のデータベースに格納された履歴情報を解析してレコメンドに必要な情報を生成する長期的嗜好分析処理装置（例えば、図１０の長期的嗜好分析処理装置４００に相当）と、前記第２のデータベースに格納されたセッションごとの履歴情報を解析してレコメンドに必要な情報を生成する短期的興味分析処理装置（例えば、図１０の短期的興味分析処理装置５００に相当）と、前記長期的嗜好分析処理装置の出力または／および前記短期的興味分析処理装置の出力を統合してレコメンド結果を出力するレコメンド結果出力装置（例えば、図１０のレコメンド結果出力装置６００に相当）と、前記レコメンド結果出力装置のレコメンド結果をユーザの興味の遷移に基づいて変更する統合割合算出装置（例えば、図１０の統合割合算出装置９００に相当）と、を備えたことを特徴とするレコメンドシステムを提案している。 (6) The present invention defines similarity between users based on user history information such as purchase history and browsing history, and recommends content that is predicted to be of interest to the user based on the defined similarity. A recommendation system that uses the collaborative filtering method that is output as a result, the first database storing all the history information (for example, corresponding to the database 200 in FIG. 10), and the first database storing the history information for each session. And a long-term preference analysis processing apparatus (for example, FIG. 10) that generates history information stored in the first database and generates information necessary for recommendation. Long-term preference analysis processing device 400), and history information for each session stored in the second database. A short-term interest analysis processing device (for example, equivalent to the short-term interest analysis processing device 500 in FIG. 10) that analyzes and generates information necessary for recommendation, and / or the output of the long-term preference analysis processing device. A recommendation result output device (for example, equivalent to the recommendation result output device 600 in FIG. 10) that integrates the output of the interest analysis processing device and outputs a recommendation result, and a recommendation result of the recommendation result output device as a transition of the user's interest A recommendation system is proposed that includes an integration rate calculation device (for example, equivalent to the integration rate calculation device 900 in FIG. 10) that is changed based on this.
　この発明によれば、第１のデータベースは、すべての履歴情報を格納する。第２のデータベースは、セッションごとの履歴情報を格納する。長期的嗜好分析処理装置は、第１のデータベースに格納された履歴情報を解析してレコメンドに必要な情報を生成する。短期的興味分析処理装置は、第２のデータベースに格納されたセッションごとの履歴情報を解析してレコメンドに必要な情報を生成する。レコメンド結果出力装置は、長期的嗜好分析処理装置の出力または／および短期的興味分析処理装置の出力を統合してレコメンド結果を出力する。統合割合算出装置は、レコメンド結果出力装置のレコメンド結果をユーザの興味の遷移に基づいて変更する。したがって、長期的嗜好分析処理装置によって、ユーザの長期的な嗜好に基づくレコメンド結果を提供し、かつ、短期的興味分析処理装置によってその状況においての短期的な興味に対するレコメンド結果を提供することで、興味の遷移を考慮した適応的なレコメンド結果を提供することが可能となる。 According to the present invention, the first database stores all history information. The second database stores history information for each session. The long-term preference analysis processing device analyzes history information stored in the first database and generates information necessary for the recommendation. The short-term interest analysis processing device analyzes the history information for each session stored in the second database and generates information necessary for the recommendation. The recommendation result output device integrates the output of the long-term preference analysis processing device and / or the output of the short-term interest analysis processing device, and outputs a recommendation result. The integrated ratio calculation device changes the recommendation result of the recommendation result output device based on the transition of the user's interest. Therefore, by providing a recommendation result based on the long-term preference of the user by the long-term preference analysis processing device, and by providing a recommendation result for the short-term interest in the situation by the short-term interest analysis processing device, It is possible to provide an adaptive recommendation result considering the transition of interest.
　（７）本発明は、（６）のレコメンドシステムについて、前記長期的嗜好分析処理装置が、すべてのログ情報の中から解析に必要な履歴情報を収集する第１の履歴情報収集手段（例えば、図１１の履歴情報収集部４０１に相当）と、該収集した履歴情報を解析して、ユーザ間の類似性を定義する第１の類似性算出手段（例えば、図１１の類似性算出部４０２に相当）と、該算出された類似性と該すべてのログ情報の中から収集された解析に必要な履歴情報とを用いて、ユーザに対する各コンテンツの推薦度を算出する第１の推薦度算出手段（例えば、図１１の推薦度算出部４０３に相当）と、該算出した推薦度を記憶する記憶手段（例えば、図１１の記憶部４０４に相当）と、を備え、前記短期的興味分析処理装置が、セッションごとのログ情報の中から解析に必要な履歴情報をリアルタイムに収集する第２の履歴情報収集手段（例えば、図１１の履歴情報収集部５０１に相当）と、該収集した履歴情報を解析して、ユーザ間の類似性を定義する第２の類似性算出手段（例えば、図１１の類似性算出部５０２に相当）と、該算出された類似性と該セッションごとのログ情報の中から解析に必要な履歴情報とを用いて、ユーザに対する各コンテンツの推薦度を算出する第２の推薦度算出手段（例えば、図１１の推薦度算出部５０３に相当）と、を備え、前記レコメンド結果出力装置が、前記統合割合算出装置の変更結果に応じて、前記記憶されたコンテンツごとの推薦度と前記第２の推薦度算出手段により算出したコンテンツごとの推薦度とからレコメンド結果を出力することを特徴とするレコメンドシステムを提案している。 (7) According to the present invention, in the recommendation system of (6), the long-term preference analysis processing device collects history information necessary for analysis from all log information (for example, first history information collection means (for example, 11) and first similarity calculation means for analyzing the collected history information and defining the similarity between users (for example, the similarity calculation unit 402 in FIG. 11). First recommendation degree calculating means for calculating the recommendation degree of each content for the user using the calculated similarity and the history information necessary for the analysis collected from all the log information (For example, equivalent to the recommendation degree calculation unit 403 in FIG. 11) and storage means (for example, equivalent to the storage unit 404 in FIG. 11) for storing the calculated recommendation degree, the short-term interest analysis processing device But for each session A second history information collecting unit (for example, corresponding to the history information collecting unit 501 in FIG. 11) that collects history information necessary for analysis from the information in real time, and analyzing the collected history information, A second similarity calculation means (for example, equivalent to the similarity calculation unit 502 in FIG. 11) that defines the similarity of the history, and the history necessary for the analysis from the calculated similarity and the log information for each session Second recommendation level calculation means (for example, equivalent to the recommendation level calculation unit 503 in FIG. 11) that calculates the recommendation level of each content for the user using the information. A recommendation result is output from the stored recommendation level for each content and the recommendation level for each content calculated by the second recommendation level calculation means according to the change result of the integrated ratio calculation device. It has proposed a recommendation system that.
　この発明によれば、長期的嗜好分析処理装置の第１の履歴情報収集手段は、すべてのログ情報の中から解析に必要な履歴情報を収集する。第１の類似性算出手段は、収集した履歴情報を解析して、ユーザ間の類似性を定義する。第１の推薦度算出手段は、算出された類似性とすべてのログ情報の中から収集された解析に必要な履歴情報とを用いて、ユーザに対する各コンテンツの推薦度を算出する。記憶手段は、算出した推薦度を記憶する。短期的興味分析処理装置の第２の履歴情報収集手段は、セッションごとのログ情報の中から解析に必要な履歴情報をリアルタイムに収集する。第２の類似性算出手段は、収集した履歴情報を解析して、ユーザ間の類似性を定義する。第２の推薦度算出手段は、算出された類似性とセッションごとのログ情報の中から解析に必要な履歴情報とを用いて、ユーザに対する各コンテンツの推薦度を算出する。レコメンド結果出力装置は、統合割合算出装置の変更結果に応じて、記憶されたコンテンツごとの推薦度と第２の推薦度算出手段により算出したコンテンツごとの推薦度とからレコメンド結果を出力する。したがって、長期的嗜好分析処理装置によって、ユーザの長期的な嗜好に基づくレコメンド結果を提供し、かつ、短期的興味分析処理装置によってその状況においての短期的な興味に対するレコメンド結果を提供することで、興味の遷移を考慮した適応的なレコメンド結果を提供することが可能となる。 According to the present invention, the first history information collecting means of the long-term preference analysis processing device collects history information necessary for analysis from all log information. The first similarity calculation unit analyzes the collected history information and defines the similarity between users. The first recommendation degree calculating means calculates the recommendation degree of each content for the user using the calculated similarity and the history information necessary for the analysis collected from all the log information. The storage means stores the calculated recommendation level. The second history information collecting means of the short-term interest analysis processing device collects history information necessary for analysis from the log information for each session in real time. The second similarity calculation means analyzes the collected history information and defines the similarity between users. The second recommendation degree calculation means calculates the recommendation degree of each content for the user using the calculated similarity and history information necessary for analysis from the log information for each session. The recommendation result output device outputs a recommendation result from the stored recommendation level for each content and the recommendation level for each content calculated by the second recommendation level calculation means according to the change result of the integrated ratio calculation device. Therefore, by providing a recommendation result based on the long-term preference of the user by the long-term preference analysis processing device, and by providing a recommendation result for the short-term interest in the situation by the short-term interest analysis processing device, It is possible to provide an adaptive recommendation result considering the transition of interest.
[規則26に基づく補充 10.01.2013]　
　（８）本発明は、（６）のレコメンドシステムについて、前記長期的嗜好分析処理装置が、すべてのログ情報の中から解析に必要な履歴情報を収集する第１の履歴情報収集手段（例えば、図１４の履歴情報収集部４０１に相当）と、該収集した履歴情報を解析して、ユーザ間の類似性を定義する第１の類似性算出手段（例えば、図１４の類似性算出部４０２に相当）と、該算出した類似性を記憶する記憶手段（例えば、図１４の記憶部４１５に相当）と、を備え、前記短期的興味分析処理装置が、セッションごとのログ情報の中から解析に必要な履歴情報をリアルタイムに収集する第２の履歴情報収集手段（例えば、図１４の履歴情報収集部５０１に相当）と、該収集した履歴情報を解析して、ユーザ間の類似性を定義する第２の類似性算出手段（例えば、図１４の類似性算出部５０２に相当）と、該セッションごとのログ情報の中から解析に必要な履歴情報と前記記憶された類似性とを用いて、ユーザに対する各コンテンツの推薦度を算出する第１の推薦度算出手段（例えば、図１４の推薦度算出部５０３に相当）と、前記第２の類似性算出手段により算出された類似性と前記セッションごとのログ情報の中から解析に必要な履歴情報とを用いて、ユーザに対する各コンテンツの推薦度を算出する第２の推薦度算出手段（例えば、図１４の推薦度算出部５１５に相当）と、を備え、前記レコメンド結果出力装置が、前記統合割合算出装置の変更結果に応じて、前記第１の推薦度算出手段により算出したコンテンツごとの推薦度と前記第２の推薦度算出手段により算出したコンテンツごとの推薦度とからレコメンド結果を出力することを特徴とするレコメンドシステムを提案している。[Supplement under rule 26 10.01.2013]
(8) In the recommendation system according to (6), the long-term preference analysis processing device collects history information necessary for analysis from all the log information (for example, first history information collection means (for example, 14) and first similarity calculation means for analyzing the collected history information and defining the similarity between users (for example, the similarity calculation unit 402 in FIG. 14). And a storage means (for example, corresponding to the storage unit 415 in FIG. 14) for storing the calculated similarity, and the short-term interest analysis processing device analyzes the log information for each session. Second history information collecting means for collecting necessary history information in real time (e.g., corresponding to the history information collection unit 501 in FIG. 14) and analyzing the collected history information to define similarity between users Second similarity calculation Each content is recommended to the user using the level (for example, equivalent to the similarity calculation unit 502 in FIG. 14), the history information necessary for analysis from the log information for each session, and the stored similarity. A first recommendation degree calculating means for calculating the degree (for example, equivalent to the recommendation degree calculating unit 503 in FIG. 14), the similarity calculated by the second similarity calculating means, and the log information for each session. Second recommendation level calculation means (for example, equivalent to the recommendation level calculation unit 515 in FIG. 14) for calculating the recommendation level of each content for the user using history information necessary for analysis from the above-mentioned recommendation The result output device determines the recommendation level for each content calculated by the first recommendation level calculation unit and the content calculated by the second recommendation level calculation unit according to the change result of the integrated ratio calculation unit. It proposes a recommendation system and outputs the recommendation result from the recommendation degree of each.
　この発明によれば、長期的嗜好分析処理装置の第１の履歴情報収集手段は、すべてのログ情報の中から解析に必要な履歴情報を収集する。第１の類似性算出手段は、収集した履歴情報を解析して、ユーザ間の類似性を定義する。記憶手段は、算出した類似性を記憶する。短期的興味分析処理装置の第２の履歴情報収集手段は、セッションごとのログ情報の中から解析に必要な履歴情報をリアルタイムに収集する。第２の類似性算出手段は、収集した履歴情報を解析して、ユーザ間の類似性を定義する。第１の推薦度算出手段は、セッションごとのログ情報の中から解析に必要な履歴情報と記憶された類似性とを用いて、ユーザに対する各コンテンツの推薦度を算出する。第２の推薦度算出手段は、第２の類似性算出手段により算出された類似性とセッションごとのログ情報の中から解析に必要な履歴情報とを用いて、ユーザに対する各コンテンツの推薦度を算出する。レコメンド結果出力装置は、統合割合算出装置の変更結果に応じて、第１の推薦度算出手段により算出したコンテンツごとの推薦度と第２の推薦度算出手段により算出したコンテンツごとの推薦度とからレコメンド結果を出力する。したがって、長期的嗜好分析処理手段によって、ユーザの長期的な嗜好に基づくレコメンド結果を提供し、かつ、短期的興味分析処理手段によってその状況においての短期的な興味に対するレコメンド結果を提供することで、興味の遷移を考慮した適応的なレコメンド結果を提供することが可能となる。 According to the present invention, the first history information collecting means of the long-term preference analysis processing device collects history information necessary for analysis from all log information. The first similarity calculation unit analyzes the collected history information and defines the similarity between users. The storage means stores the calculated similarity. The second history information collecting means of the short-term interest analysis processing device collects history information necessary for analysis from the log information for each session in real time. The second similarity calculation means analyzes the collected history information and defines the similarity between users. The first recommendation level calculation means calculates the recommendation level of each content for the user by using the history information necessary for the analysis and the stored similarity from the log information for each session. The second recommendation level calculation means uses the similarity calculated by the second similarity calculation means and the history information necessary for analysis from the log information for each session to determine the recommendation level of each content to the user. calculate. The recommendation result output device uses the recommendation level for each content calculated by the first recommendation level calculation unit and the recommendation level for each content calculated by the second recommendation level calculation unit according to the change result of the integrated ratio calculation device. Output recommendation results. Therefore, by providing a recommendation result based on the long-term preference of the user by the long-term preference analysis processing means, and by providing a recommendation result for the short-term interest in the situation by the short-term interest analysis processing means, It is possible to provide an adaptive recommendation result considering the transition of interest.
　（９）本発明は、（６）のレコメンドシステムについて、前記長期的嗜好分析処理装置が、すべてのログ情報の中から解析に必要な履歴情報を収集する第１の履歴情報収集手段（例えば、図１７の履歴情報収集部４０１に相当）と、該収集した履歴情報を解析して、ユーザ間の類似性を定義する第１の類似性算出手段（例えば、図１７の類似性算出部４０２に相当）と、該算出された類似性該セッションごとのログ情報の中から解析に必要な履歴情報とを用いて、ユーザに対する各コンテンツの推薦度を算出する第１の推薦度算出手段（例えば、図１７の推薦度算出部４０３に相当）と、該定義した類似性および算出した推薦度を記憶する記憶手段（例えば、図１７の記憶部４２６に相当）と、を備え、前記短期的興味分析処理装置が、セッションごとのログ情報の中から解析に必要な履歴情報をリアルタイムに収集する第２の履歴情報収集手段（例えば、図１７の履歴情報収集部５０１に相当）と、前記定義した類似性および算出した推薦度と前記セッションごとのログ情報の中から解析に必要な履歴情報とを用いて、ユーザに対する各コンテンツの推薦度を算出する第２の推薦度算出手段（例えば、図１７の推薦度算出部５２６に相当）と、前記レコメンド結果出力装置が、前記統合割合算出装置の変更結果に応じて、前記第１の推薦度算出手段により算出したコンテンツごとの推薦度と前記第２の推薦度算出手段により算出したコンテンツごとの推薦度とからレコメンド結果を出力することを特徴とするレコメンドシステムを提案している。 (9) In the recommendation system according to (6), the long-term preference analysis processing device collects history information necessary for analysis from all the log information (for example, first history information collection means (for example, 17) and first similarity calculation means for analyzing the collected history information and defining the similarity between users (for example, the similarity calculation unit 402 in FIG. 17). First recommendation degree calculating means (for example, calculating the recommendation degree of each content for the user, using the calculated similarity and the history information necessary for analysis from the log information for each session) 17), and a storage means (for example, corresponding to the storage unit 426 in FIG. 17) for storing the defined similarity and the calculated recommendation level. The processing unit Second history information collection means (for example, corresponding to the history information collection unit 501 in FIG. 17) that collects history information necessary for analysis from the log information for each session in real time, and the similarity defined and calculated Second recommendation degree calculation means for calculating the recommendation degree of each content for the user using the recommendation degree and the history information necessary for analysis from the log information for each session (for example, the recommendation degree calculation unit in FIG. 17). And the recommendation result output device according to the change result of the integrated ratio calculation device, the recommendation level for each content calculated by the first recommendation level calculation unit, and the second recommendation level calculation unit. The recommendation system is characterized in that the recommendation result is output from the recommendation level calculated for each content.
　この発明によれば、長期的嗜好分析処理装置の第１の履歴情報収集手段は、すべてのログ情報の中から解析に必要な履歴情報を収集する。第１の類似性算出手段は、収集した履歴情報を解析して、ユーザ間の類似性を定義する。第１の推薦度算出手段は、算出された類似性とセッションごとのログ情報の中から解析に必要な履歴情報とを用いて、ユーザに対する各コンテンツの推薦度を算出する。記憶手段は、定義した類似性および算出した推薦度を記憶する。短期的興味分析処理装置の第２の履歴情報収集手段は、セッションごとのログ情報の中から解析に必要な履歴情報をリアルタイムに収集する。第２の推薦度算出手段は、定義した類似性および算出した推薦度とセッションごとのログ情報の中から解析に必要な履歴情報とを用いて、ユーザに対する各コンテンツの推薦度を算出する。レコメンド結果出力装置は、統合割合算出装置の変更結果に応じて、第１の推薦度算出手段により算出したコンテンツごとの推薦度と第２の推薦度算出手段により算出したコンテンツごとの推薦度とからレコメンド結果を出力する。したがって、長期的嗜好分析処理装置によって、ユーザの長期的な嗜好に基づくレコメンド結果を提供し、かつ、短期的興味分析処理装置によってその状況においての短期的な興味に対するレコメンド結果を提供することで、興味の遷移を考慮した適応的なレコメンド結果を提供することが可能となる。 According to the present invention, the first history information collecting means of the long-term preference analysis processing device collects history information necessary for analysis from all log information. The first similarity calculation unit analyzes the collected history information and defines the similarity between users. The first recommendation level calculating means calculates the recommendation level of each content for the user using the calculated similarity and history information necessary for analysis from the log information for each session. The storage means stores the defined similarity and the calculated recommendation level. The second history information collecting means of the short-term interest analysis processing device collects history information necessary for analysis from the log information for each session in real time. The second recommendation degree calculation means calculates the recommendation degree of each content for the user using the defined similarity, the calculated recommendation degree, and history information necessary for analysis from the log information for each session. The recommendation result output device uses the recommendation level for each content calculated by the first recommendation level calculation unit and the recommendation level for each content calculated by the second recommendation level calculation unit according to the change result of the integrated ratio calculation device. Output recommendation results. Therefore, by providing a recommendation result based on the long-term preference of the user by the long-term preference analysis processing device, and by providing a recommendation result for the short-term interest in the situation by the short-term interest analysis processing device, It is possible to provide an adaptive recommendation result considering the transition of interest.
[規則26に基づく補充 10.01.2013]　
　（１０）本発明は、（６）から（９）のレコメンドシステムについて、出力するレコメンドについて、ユーザの閲覧状況を取得する閲覧状況取得装置と、該取得したユーザの閲覧状況を前記第２のデータベースと前記統合割合算出装置にフィードバックする閲覧状況フィードバック装置（例えば、図１０の閲覧状況フィードバック装置８００に相当）と、を備え、前記統合割合算出装置が閲覧状況フィードバック情報に基づいて、前記レコメンド結果出力装置のレコメンド結果をユーザの興味の遷移に基づいて変更することを特徴とするレコメンドシステムを提案している。[Supplement under rule 26 10.01.2013]
(10) The present invention relates to the recommendation system of (6) to (9), a browsing status acquisition apparatus that acquires the browsing status of the user for the recommendations to be output, and the acquired browsing status of the user in the second database. And a browsing status feedback device (for example, corresponding to the browsing status feedback device 800 of FIG. 10) that feeds back to the integrated ratio calculating device, and the integrated ratio calculating device outputs the recommendation result based on the browsing status feedback information The recommendation system characterized by changing the recommendation result of an apparatus based on transition of a user's interest is proposed.
　この発明によれば、閲覧状況取得装置は、レコメンド結果出力装置から出力されるレコメンドについて、ユーザの閲覧状況を取得する。閲覧状況フィードバック装置は、取得したユーザの閲覧状況をデータベースと統合割合算出装置にフィードバックする。統合割合算出装置は、閲覧状況フィードバック情報に基づいて、レコメンド結果出力装置のレコメンド結果をユーザの興味の遷移に基づいて変更する。したがって、取得したユーザの閲覧状況をデータベースと統合割合算出装置にフィードバックし、統合割合算出装置が、閲覧状況フィードバック情報に基づいて、レコメンド結果出力装置のレコメンド結果をユーザの興味の遷移に基づいて変更するため、ユーザが短期的に興味をもったコンテンツ等を的確に抽出して、レコメンドを行うことができる。 According to the present invention, the browsing status acquisition device acquires the browsing status of the user for the recommendations output from the recommendation result output device. The browsing status feedback device feeds back the acquired browsing status of the user to the database and the integrated ratio calculation device. The integrated ratio calculation device changes the recommendation result of the recommendation result output device based on the transition of the user's interest based on the browsing status feedback information. Therefore, the obtained browsing status of the user is fed back to the database and the integrated rate calculating device, and the integrated rate calculating device changes the recommended result of the recommended result output device based on the transition of the user's interest based on the browsing status feedback information. Therefore, it is possible to make a recommendation by accurately extracting contents and the like that the user is interested in in the short term.
　（１１）本発明は、購買履歴や閲覧履歴等のユーザの履歴情報を元にユーザ間の類似性を定義し、該定義した類似性に基づいてユーザが興味を持つと予測されるコンテンツをレコメンド結果として出力する協調フィルタリング方法を用いるレコメンド方法であって、すべての前記履歴情報を解析してレコメンドに必要な情報を生成する長期的嗜好分析処理ステップと、セッションごとの履歴情報を解析してレコメンドに必要な情報を生成する短期的興味分析処理ステップと、前記長期的嗜好分析処理ステップの出力または／および前記短期的興味分析処理ステップの出力を統合してレコメンド結果を出力するレコメンド結果出力ステップと、前記レコメンド結果出力ステップのレコメンド結果をユーザの興味の遷移に基づいて変更する統合割合算出ステップと、を備えたことを特徴とするレコメンド方法を提案している。 (11) The present invention defines similarity between users based on user history information such as purchase history and browsing history, and recommends content predicted to be of interest to the user based on the defined similarity. It is a recommendation method using a collaborative filtering method that is output as a result, and a long-term preference analysis processing step that analyzes all the history information to generate information necessary for the recommendation, and analyzes the history information for each session to recommend A short-term interest analysis processing step for generating necessary information, and a recommendation result output step for outputting a recommendation result by integrating the output of the long-term preference analysis processing step and / or the output of the short-term interest analysis processing step; , Integration to change the recommendation result of the recommendation result output step based on the transition of the user's interest It proposes a recommendation method characterized by comprising: a case calculating step.
　この発明によれば、すべての履歴情報を解析してレコメンドに必要な情報を生成するとともに、セッションごとの履歴情報を解析してレコメンドに必要な情報を生成する。そして、上記２つの処理出力を統合してレコメンド結果を出力し、レコメンド結果をユーザの興味の遷移に基づいて変更する。したがって、長期的嗜好分析処理ステップによって、ユーザの長期的な嗜好に基づくレコメンド結果を提供し、かつ、短期的興味分析処理ステップによってその状況においての短期的な興味に対するレコメンド結果を提供することで、興味の遷移を考慮した適応的なレコメンド結果を提供することが可能となる。 According to the present invention, all the history information is analyzed to generate information necessary for the recommendation, and the history information for each session is analyzed to generate information necessary for the recommendation. Then, the above two processing outputs are integrated to output a recommendation result, and the recommendation result is changed based on the transition of the user's interest. Therefore, by providing a recommendation result based on the long-term preference of the user by the long-term preference analysis processing step, and by providing a recommendation result for the short-term interest in the situation by the short-term interest analysis processing step, It is possible to provide an adaptive recommendation result considering the transition of interest.
　（１２）本発明は、購買履歴や閲覧履歴等のユーザの履歴情報を元にユーザ間の類似性を定義し、該定義した類似性に基づいてユーザが興味を持つと予測されるコンテンツをレコメンド結果として出力する協調フィルタリング方法を用いるレコメンド方法をコンピュータに実行させるためのプログラムであって、すべての前記履歴情報を解析してレコメンドに必要な情報を生成する長期的嗜好分析処理ステップと、セッションごとの履歴情報を解析してレコメンドに必要な情報を生成する短期的興味分析処理ステップと、前記長期的嗜好分析処理ステップの出力または／および前記短期的興味分析処理ステップの出力を統合してレコメンド結果を出力するレコメンド結果出力ステップと、前記レコメンド結果出力ステップのレコメンド結果をユーザの興味の遷移に基づいて変更する統合割合算出ステップと、をコンピュータに実行させるためのプログラムを提案している。 (12) The present invention defines similarity between users based on user history information such as purchase history and browsing history, and recommends content predicted to be of interest to the user based on the defined similarity. A program for causing a computer to execute a recommendation method using a collaborative filtering method to be output as a result, analyzing all the history information and generating information necessary for the recommendation, and a session Results of recommendation by integrating the output of the short-term interest analysis processing step and / or the output of the long-term preference analysis processing step and / or the output of the short-term interest analysis processing step of analyzing the history information of A recommendation result output step, and a recommendation result of the recommendation result output step We propose a program for executing the integrated ratio calculating step of changing on the basis of the transition of the user's interest, to a computer.
　この発明によれば、すべての履歴情報を解析してレコメンドに必要な情報を生成するとともに、セッションごとの履歴情報を解析してレコメンドに必要な情報を生成する。そして、上記２つの処理出力を統合してレコメンド結果を出力し、レコメンド結果をユーザの興味の遷移に基づいて変更する。したがって、長期的嗜好分析処理ステップによって、ユーザの長期的な嗜好に基づくレコメンド結果を提供し、かつ、短期的興味分析処理ステップによってその状況においての短期的な興味に対するレコメンド結果を提供することで、興味の遷移を考慮した適応的なレコメンド結果を提供することが可能となる。 According to the present invention, all the history information is analyzed to generate information necessary for the recommendation, and the history information for each session is analyzed to generate information necessary for the recommendation. Then, the above two processing outputs are integrated to output a recommendation result, and the recommendation result is changed based on the transition of the user's interest. Therefore, by providing a recommendation result based on the long-term preference of the user by the long-term preference analysis processing step, and by providing a recommendation result for the short-term interest in the situation by the short-term interest analysis processing step, It is possible to provide an adaptive recommendation result considering the transition of interest.
　本発明によれば、長期的嗜好分析処理によって、ユーザの長期的な嗜好に基づくレコメンド結果を提供し、かつ、短期的興味分析処理によってその状況においての短期的な興味に対するレコメンド結果を提供することで、興味の遷移を考慮した適応的なレコメンド結果を提供することが可能となるという効果がある。 According to the present invention, a recommendation result based on a user's long-term preference is provided by a long-term preference analysis process, and a recommendation result for a short-term interest in the situation is provided by a short-term interest analysis process. Thus, there is an effect that it is possible to provide an adaptive recommendation result considering the transition of interest.
本発明の第１の実施形態に係るレコメンド装置の構成を示す図である。It is a figure which shows the structure of the recommendation apparatus which concerns on the 1st Embodiment of this invention. 本発明の第１の実施形態に係るレコメンド装置の詳細な構成を示す図である。It is a figure which shows the detailed structure of the recommendation apparatus which concerns on the 1st Embodiment of this invention. 本発明の第１の実施形態に係るレコメンド装置の処理を示す図である。It is a figure which shows the process of the recommendation apparatus which concerns on the 1st Embodiment of this invention. 本発明の第２の実施形態に係るレコメンド装置の構成を示す図である。It is a figure which shows the structure of the recommendation apparatus which concerns on the 2nd Embodiment of this invention. 本発明の第２の実施形態に係るレコメンド装置の詳細な構成を示す図である。It is a figure which shows the detailed structure of the recommendation apparatus which concerns on the 2nd Embodiment of this invention. 本発明の第２の実施形態に係るレコメンド装置の処理を示す図である。It is a figure which shows the process of the recommendation apparatus which concerns on the 2nd Embodiment of this invention. 本発明の第３の実施形態に係るレコメンド装置の構成を示す図である。It is a figure which shows the structure of the recommendation apparatus which concerns on the 3rd Embodiment of this invention. 本発明の第３の実施形態に係るレコメンド装置の詳細な構成を示す図である。It is a figure which shows the detailed structure of the recommendation apparatus which concerns on the 3rd Embodiment of this invention. 本発明の第３の実施形態に係るレコメンド装置の処理を示す図である。It is a figure which shows the process of the recommendation apparatus which concerns on the 3rd Embodiment of this invention. 本発明の第４の実施形態に係るレコメンドシステムの構成を示す図である。It is a figure which shows the structure of the recommendation system which concerns on the 4th Embodiment of this invention. 本発明の第４の実施形態に係るレコメンドシステムの詳細な構成を示す図である。It is a figure which shows the detailed structure of the recommendation system which concerns on the 4th Embodiment of this invention. 本発明の第４の実施形態に係るレコメンドシステムの処理を示す図である。It is a figure which shows the process of the recommendation system which concerns on the 4th Embodiment of this invention. 本発明の第５の実施形態に係るレコメンドシステムの構成を示す図である。It is a figure which shows the structure of the recommendation system which concerns on the 5th Embodiment of this invention. 本発明の第５の実施形態に係るレコメンドシステムの詳細な構成を示す図である。It is a figure which shows the detailed structure of the recommendation system which concerns on the 5th Embodiment of this invention. 本発明の第５の実施形態に係るレコメンドシステムの処理を示す図である。It is a figure which shows the process of the recommendation system which concerns on the 5th Embodiment of this invention. 本発明の第６の実施形態に係るレコメンドシステムの構成を示す図である。It is a figure which shows the structure of the recommendation system which concerns on the 6th Embodiment of this invention. 本発明の第６の実施形態に係るレコメンドシステムの詳細な構成を示す図である。It is a figure which shows the detailed structure of the recommendation system which concerns on the 6th Embodiment of this invention. 本発明の第６の実施形態に係るレコメンドシステムの処理を示す図である。It is a figure which shows the process of the recommendation system which concerns on the 6th Embodiment of this invention.
　以下、本発明の実施形態について、図面を用いて、詳細に説明する。
　なお、本実施形態における構成要素は適宜、既存の構成要素等との置き換えが可能であり、また、他の既存の構成要素との組合せを含む様々なバリエーションが可能である。したがって、本実施形態の記載をもって、特許請求の範囲に記載された発明の内容を限定するものではない。 Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.
Note that the constituent elements in the present embodiment can be appropriately replaced with existing constituent elements and the like, and various variations including combinations with other existing constituent elements are possible. Therefore, the description of the present embodiment does not limit the contents of the invention described in the claims.
＜第１の実施形態＞
　図１および図３を用いて、本発明の第１の実施形態について説明する。 <First Embodiment>
A first embodiment of the present invention will be described with reference to FIGS. 1 and 3.
＜レコメンド装置の構成＞
　図１を用いて、本実施形態に係るレコメンド装置の構成について説明する。 <Configuration of recommendation device>
The configuration of the recommendation device according to the present embodiment will be described with reference to FIG.
　本実施形態に係るレコメンド装置１０００は、図１に示すように、長期的嗜好分析処理部１１００と、短期的興味分析処理部１２００と、レコメンド結果出力部１３００と、レコメンド表示部１４００と、閲覧状況フィードバック部１５００と、統合割合算出部１６００とから構成され、長期的嗜好分析処理部１１００には、データベース２００が、短期的興味分析処理部１２００には、データベース３００が接続されている。 As shown in FIG. 1, the recommendation device 1000 according to the present embodiment includes a long-term preference analysis processing unit 1100, a short-term interest analysis processing unit 1200, a recommendation result output unit 1300, a recommendation display unit 1400, and a browsing situation. The database 200 is connected to the long-term preference analysis processing unit 1100, and the database 300 is connected to the short-term interest analysis processing unit 1200.
　なお、データベース２００には、すべてのユーザのすべてのログデータ、例えば、商品の購入履歴やウェブページの閲覧履歴を記録した、全データベースが格納されている。一方、データベース３００には、短期的興味分析処理を行うため、セッションごとのログデータが格納されている。具体的には、上記のデータベースより、例えば、ウェブページの閲覧履歴においては、各ユーザの３０分以内の連続した閲覧履歴を一セッションとして履歴を細分化したもの、あるいは、それらからランダムに選択したもの、または、各セッション間において共起する二つのウェブページ組をカウントし、それが別途設けた閾値を越えたウェブページ組を関連のあるウェブページとみなして、関連のある複数のウェブページを履歴としたものを、閲覧履歴とみなすデータベースを作成する。ただし、データ縮約手法についてはこの手法に限るものではない。 The database 200 stores all the log data of all users, for example, all the databases that record the purchase history of products and the browsing history of web pages. On the other hand, the database 300 stores log data for each session in order to perform short-term interest analysis processing. Specifically, from the above database, for example, in the browsing history of web pages, each user's continuous browsing history within 30 minutes is subdivided into one session, or randomly selected from them. Or two web page sets that co-occur between each session, and the web page set that exceeds the separately set threshold is regarded as a related web page, and a plurality of related web pages are Create a database that considers history as browsing history. However, the data reduction method is not limited to this method.
　また、長期的嗜好分析処理部１１００においては、全データベースより被推薦ユーザおよびそのほかの全ユーザの履歴情報を取得し、被推薦ユーザと他のユーザ間との類似性を類似度として出力し、上記の出力と全データベースの履歴情報より被推薦ユーザに対するひとつひとつのコンテンツのお勧めの度合を推薦度として出力し、上記出力を保存する。以上の処理を、被推薦ユーザを変更し、全ユーザを対象として行う。 Further, the long-term preference analysis processing unit 1100 acquires history information of the recommended user and all other users from all databases, outputs the similarity between the recommended user and other users as the similarity, The degree of recommendation of each content for the recommended user is output as a recommendation degree from the output of the above and the history information of all databases, and the output is stored. The above processing is performed for all users by changing the recommended user.
　また、短期的興味分析処理部１２００においては、縮約されたデータベースより、上記と同様に、３０分以内の連続した閲覧中のウェブページをセッションとし、そのセッション中に含まれるコンテンツを当該ユーザの履歴とし、被推薦ユーザと他のユーザ間との類似性を類似度として出力し、上記の出力と縮約されたデータベースの履歴情報より被推薦ユーザに対するひとつひとつのコンテンツのお勧めの度合を推薦度として出力する。 Further, in the short-term interest analysis processing unit 1200, from the contracted database, similarly to the above, a web page being continuously browsed within 30 minutes is set as a session, and the content included in the session is set as the user's content. As a history, the similarity between the recommended user and other users is output as the similarity, and the recommended degree of the recommended content of each content for the recommended user from the above output and the history information of the reduced database Output as.
　レコメンド結果出力部１３００では、長期的嗜好分析処理部１１００で保存された記録より、被推薦ユーザに対するひとつひとつのコンテンツの推薦度を取得する。一方で、短期的興味分析処理部１２００による出力を同時に取得し、ひとつひとつのコンテンツの推薦度を、例えば、二つの推薦度の加重平均をとるなど、統合割合算出部１６００により定められた方法によって最終的な推薦度を算出する。 The recommendation result output unit 1300 acquires the recommendation level of each content for the recommended user from the record stored in the long-term preference analysis processing unit 1100. On the other hand, the output from the short-term interest analysis processing unit 1200 is acquired at the same time, and the recommendation level of each content is finalized by a method determined by the integration ratio calculation unit 1600, for example, taking a weighted average of the two recommendation levels. The degree of recommendation is calculated.
　ここで、統合割合算出部１６００は、当該ユーザのセッションのコンテンツ数Ｎｕｍ（ｓｅｓｓｉｏｎ）や、閲覧時間Ｔｉｍｅ（ｓｅｓｓｉｏｎ）を考慮し、例えば、以下の数式（１）、（２）、（３）や、それらの組み合わせ等によって長期的嗜好分析処理による各コンテンツの推薦度Ｒｅｃ（ｂａｔｃｈ）と短期的興味分析処理による各コンテンツの推薦度Ｒｅｃ（ｒｅａｌ）を統合し、最終的な推薦度Ｒｅｃ(Ｆｉｎａｌ)を算出する。
（１）　α＝γ＾（Ｎｕｍ（ｓｅｓｓｉｏｎ））、（γは０＜γ＜１を満たす定数）
（２）　α＝γ＾（Ｎｕｍ（ｓｅｓｓｉｏｎ）／Ｔｉｍｅ（ｓｅｓｓｉｏｎ））、（γは０＜γ＜１を満たす定数）
（３）　α＝１／（１＋ｅｘｐ（a－ｂ＊Ｎｕｍ（ｓｅｓｓｉｏｎ）））
ただし、Ｒｅｃ（Ｆｉｎａｌ）＝αＲｅｃ（ｂａｔｃｈ）＋（1－α）Ｒｅｃ（ｒｅａｌ）。
（（３）は、ロジスティック式を意味する。） Here, the integration ratio calculation unit 1600 takes into account the number of contents of the user's session Num (session) and the browsing time Time (session), for example, the following mathematical formulas (1), (2), (3) The recommendation level Rec (batch) of each content by the long-term preference analysis process and the recommendation level Rec (real) of each content by the short-term interest analysis process are integrated by combining them, and the final recommendation level Rec (Final) Is calculated.
(1) α = γ ^ (Num (session)), (γ is a constant satisfying 0 <γ <1)
(2) α = γ ^ (Num (session) / Time (session)), (γ is a constant satisfying 0 <γ <1)
(3) α = 1 / (1 + exp (ab−Num (session)))
However, Rec (Final) = αRec (batch) + (1−α) Rec (real).
((3) means logistic formula.)
＜レコメンド装置の詳細な構成＞
　本実施形態に係るレコメンド装置は、図２に示すように、長期的嗜好分析処理部１１００を構成する履歴情報収集部１１０１と、類似性算出部１１０２と、推薦度算出部１１０３と、記憶部１１０４と、短期的興味分析処理部１２００を構成する履歴情報収集部１２０１と、類似性算出部１２０２と、推薦度算出部１２０３と、レコメンド結果出力部１３００、レコメンド表示部１４００、閲覧状況フィードバック部１５００、統合割合算出部１６００とから構成されている。 <Detailed configuration of recommendation device>
As illustrated in FIG. 2, the recommendation device according to the present embodiment includes a history information collection unit 1101, a similarity calculation unit 1102, a recommendation degree calculation unit 1103, and a storage unit 1104 that constitute a long-term preference analysis processing unit 1100. A history information collection unit 1201, a similarity calculation unit 1202, a recommendation degree calculation unit 1203, a recommendation result output unit 1300, a recommendation display unit 1400, a browsing status feedback unit 1500, which constitute the short-term interest analysis processing unit 1200. And an integrated ratio calculation unit 1600.
　履歴情報収集部１１０１は、すべてのログ情報の中から解析に必要な履歴情報をデータベース２００から収集する。類似性算出部１１０２は、収集した履歴情報を解析して、ユーザ間の類似性を定義する。推薦度算出部１１０３は、算出された類似性とすべてのログ情報の中から収集された解析に必要な履歴情報とを用いて、ユーザに対する各コンテンツの推薦度を算出する。記憶部１１０４は、算出した推薦度を記憶する。 The history information collection unit 1101 collects history information necessary for analysis from all the log information from the database 200. The similarity calculation unit 1102 analyzes the collected history information and defines the similarity between users. The recommendation level calculation unit 1103 calculates the recommendation level of each content for the user by using the calculated similarity and history information necessary for analysis collected from all log information. The storage unit 1104 stores the calculated recommendation level.
　履歴情報収集部１２０１は、セッションごとのログ情報の中から解析に必要な履歴情報をデータベース３００リアルタイムに収集する。類似性算出部１２０２は、収集した履歴情報を解析して、ユーザ間の類似性を定義する。推薦度算出部１２０３は、算出された類似性とセッションごとのログ情報の中から解析に必要な履歴情報とを用いて、ユーザに対する各コンテンツの推薦度を算出する。 The history information collection unit 1201 collects history information necessary for analysis from the log information for each session in real time in the database 300. The similarity calculation unit 1202 analyzes the collected history information and defines the similarity between users. The recommendation level calculation unit 1203 calculates the recommendation level of each content for the user using the calculated similarity and history information necessary for analysis from the log information for each session.
　レコメンド結果出力部１３００は、統合割合算出部１６００の変更結果に応じて、記憶されたコンテンツごとの推薦度と推薦度算出部１２０３により算出したコンテンツごとの推薦度とからレコメンド結果を出力する。なお、統合割合算出部１６００の機能は、加重和を変化させることであり、例えば、セッションの閲覧個数やセッションタイム等に基づいて、指数・ロジスティック関数等により処理される。 The recommendation result output unit 1300 outputs a recommendation result from the stored recommendation level for each content and the recommendation level for each content calculated by the recommendation level calculation unit 1203 according to the change result of the integration ratio calculation unit 1600. Note that the function of the integration ratio calculation unit 1600 is to change the weighted sum, and is processed by an index / logistic function or the like based on, for example, the number of sessions viewed or the session time.
＜レコメンド装置の処理＞
　図３を用いて、本実施形態に係るレコメンド装置の処理について説明する。 <Processing of recommendation device>
The processing of the recommendation device according to the present embodiment will be described using FIG.
　まず、長期的嗜好分析処理部１１００の履歴情報収集部１１０１は、すべてのログ情報の中から解析に必要な履歴情報をデータベース２００から収集し（ステップＳ１０１）、類似性算出部１１０２は、収集した履歴情報を解析して、ユーザ間の類似性を、例えば、二人のユーザの履歴を比較して同時に履歴を持つコンテンツ数を元に、または、二つのコンテンツの履歴を比較して同時に履歴を持つコンテンツ数を元に定義する（ステップＳ１０２）。推薦度算出部１１０３は、算出された類似性とすべてのログ情報の中から解析に必要な履歴情報とを用いて、ユーザに対する各コンテンツの推薦度を、例えば、当該ユーザのコンテンツの履歴を持つ他のユーザとの類似性、または、当該ユーザの履歴に存在するコンテンツと他のコンテンツとの類似性を元に算出し（ステップＳ１０３）、記憶部１１０４は、算出した推薦度を記憶する（ステップＳ１０４）。 First, the history information collection unit 1101 of the long-term preference analysis processing unit 1100 collects history information necessary for analysis from all the log information from the database 200 (step S101), and the similarity calculation unit 1102 collects the history information. Analyze history information and compare similarities between users, for example, compare the history of two users based on the number of content with history at the same time, or compare the history of two content and The definition is made based on the number of contents possessed (step S102). The recommendation level calculation unit 1103 uses the calculated similarity and history information necessary for analysis from all the log information to determine the recommendation level of each content for the user, for example, the content history of the user. Calculation is made based on the similarity with other users or the similarity between the contents existing in the user's history and other contents (step S103), and the storage unit 1104 stores the calculated recommendation degree (step S103). S104).
　短期的興味分析処理部１２００の履歴情報収集部１２０１は、セッションごとのログ情報の中から解析に必要な履歴情報をデータベース３００から収集し（ステップＳ１０５）、類似性算出部１２０２は、収集した履歴情報を解析して、ユーザ間の類似性を、前述同様に定義する（ステップＳ１０６）。推薦度算出部１２０３は、算出された類似性とセッションごとのログ情報の中から解析に必要な履歴情報とを用いて、ユーザに対する各コンテンツの推薦度を算出する（ステップＳ１０７）。 The history information collection unit 1201 of the short-term interest analysis processing unit 1200 collects history information necessary for analysis from the log information for each session (step S105), and the similarity calculation unit 1202 collects the collected history information. By analyzing the information, the similarity between users is defined as described above (step S106). The recommendation level calculation unit 1203 calculates the recommendation level of each content for the user using the calculated similarity and history information necessary for analysis from the log information for each session (step S107).
　そして、レコメンド結果出力部１３００は、統合割合算出部１６００の変更結果に応じて、記憶されたコンテンツごとの推薦度と推薦度算出部１２０３により算出したコンテンツごとの推薦度とからレコメンド結果を出力する（ステップＳ１０８）。 Then, the recommendation result output unit 1300 outputs a recommendation result from the stored recommendation level for each content and the recommendation level for each content calculated by the recommendation level calculation unit 1203 according to the change result of the integration ratio calculation unit 1600. (Step S108).
　以上、説明したように、本実施形態によれば、長期的嗜好分析処理において算出した類似性を用いた推薦度と短期的興味分析処理において算出した類似性を用いた推薦度とからレコメンド結果を出力するため、長期的嗜好分析処理部によって、ユーザの長期的な嗜好に基づくレコメンド結果を提供し、かつ、短期的興味分析処理部によってその状況においての短期的な興味に対するレコメンド結果を提供することにより、興味の遷移を考慮した適応的なレコメンド結果を提供することが可能となる。 As described above, according to the present embodiment, the recommendation result is obtained from the recommendation degree using the similarity calculated in the long-term preference analysis process and the recommendation degree using the similarity calculated in the short-term interest analysis process. To provide a recommendation result based on the user's long-term preference by the long-term preference analysis processing unit, and to provide a recommendation result for the short-term interest in the situation by the short-term interest analysis processing unit. Thus, it is possible to provide an adaptive recommendation result considering the transition of interest.
＜第２の実施形態＞
　図４から図６を用いて、本発明の第２の実施形態について説明する。 <Second Embodiment>
A second embodiment of the present invention will be described with reference to FIGS.
＜レコメンド装置の構成＞
　図４を用いて、本実施形態に係るレコメンド装置の構成について説明する。 <Configuration of recommendation device>
The configuration of the recommendation device according to the present embodiment will be described with reference to FIG.
　本実施形態に係るレコメンド装置１０００は、図４に示すように、長期的嗜好分析処理部１１１０と、短期的興味分析処理部１２１０と、レコメンド結果出力部１３１０と、レコメンド表示部１４００と、閲覧状況フィードバック部１５００と、統合割合算出部１６００とから構成され、長期的嗜好分析処理部１１１０には、データベース２００が、短期的興味分析処理部１２１０には、データベース３００が接続されている。 As shown in FIG. 4, the recommendation device 1000 according to the present embodiment includes a long-term preference analysis processing unit 1110, a short-term interest analysis processing unit 1210, a recommendation result output unit 1310, a recommendation display unit 1400, and a browsing situation. The database 200 is connected to the long-term preference analysis processing unit 1110, and the database 300 is connected to the short-term interest analysis processing unit 1210.
＜レコメンド装置の詳細な構成＞
　本実施形態に係るレコメンド装置は、図５に示すように、長期的嗜好分析処理部１１１０を構成する履歴情報収集部１１０１と、類似性算出部１１０２と、記憶部１１１５と、短期的興味分析処理部１２１０を構成する履歴情報収集部１２０１と、類似性算出部１２０２と、推薦度算出部１２０３と、推薦度算出部１２１５と、レコメンド結果出力部１３１０と、レコメンド表示部１４００と、閲覧状況フィードバック部１５００と、統合割合算出部１６００とから構成されている。なお、第１の実施形態と同様の符号を付す構成要素については、同様の機能を有することから、その詳細な説明は、省略する。 <Detailed configuration of recommendation device>
As shown in FIG. 5, the recommendation device according to the present embodiment includes a history information collection unit 1101, a similarity calculation unit 1102, a storage unit 1115, and a short-term interest analysis process that constitute a long-term preference analysis processing unit 1110. History information collection unit 1201, similarity calculation unit 1202, recommendation level calculation unit 1203, recommendation level calculation unit 1215, recommendation result output unit 1310, recommendation display unit 1400, and browsing status feedback unit that constitute unit 1210 1500 and an integration ratio calculation unit 1600. In addition, about the component which attaches | subjects the code | symbol similar to 1st Embodiment, since it has the same function, the detailed description is abbreviate | omitted.
　記憶部１１１５は、類似性算出部１１０２が算出した類似性を記憶する。推薦度算出部１２１５は、記憶部１１１５に記憶されている類似性と、セッションごとのログ情報の中から解析に必要な履歴情報とを用いて、ユーザに対する各コンテンツの推薦度を算出する。 The storage unit 1115 stores the similarity calculated by the similarity calculation unit 1102. The recommendation level calculation unit 1215 calculates the recommendation level of each content for the user using the similarity stored in the storage unit 1115 and the history information necessary for analysis from the log information for each session.
　レコメンド結果出力部１３１０は、統合割合算出部１６００の変更結果に応じて、推薦度算出部１２０３により算出したコンテンツごとの推薦度と推薦度算出部１２１５により算出したコンテンツごとの推薦度とからレコメンド結果を出力する。 The recommendation result output unit 1310 determines a recommendation result from the recommendation level for each content calculated by the recommendation level calculation unit 1203 and the recommendation level for each content calculated by the recommendation level calculation unit 1215 according to the change result of the integration ratio calculation unit 1600. Is output.
＜レコメンド装置の処理＞
　図６を用いて、本実施形態に係るレコメンド装置の処理について説明する。 <Processing of recommendation device>
The process of the recommendation apparatus according to the present embodiment will be described with reference to FIG.
　まず、長期的嗜好分析処理部１１１０の履歴情報収集部１１０１は、すべてのログ情報の中から解析に必要な履歴情報をデータベース２００から収集し（ステップＳ２０１）、類似性算出部１１０２は、収集した履歴情報を解析して、ユーザ間の類似性を定義し（ステップＳ２０２）、記憶部１１１５は、算出した類似性を記憶する（ステップＳ２０３）。 First, the history information collection unit 1101 of the long-term preference analysis processing unit 1110 collects history information necessary for analysis from all log information from the database 200 (step S201), and the similarity calculation unit 1102 collects the history information. The history information is analyzed to define the similarity between users (step S202), and the storage unit 1115 stores the calculated similarity (step S203).
　短期的興味分析処理部１２１０の履歴情報収集部１２０１は、セッションごとのログ情報の中から解析に必要な履歴情報をデータベース３００からリアルタイム収集し（ステップＳ２０４）、類似性算出部１２０２は、収集した履歴情報を解析して、ユーザ間の類似性を定義する（ステップＳ２０５）。推薦度算出部１２０３は、算出された類似性とセッションごとのログ情報の中から解析に必要な履歴情報とを用いて、ユーザに対する各コンテンツの推薦度を算出する（ステップＳ２０６）。また、推薦度算出部１２１５は、類似性算出部１１０２により算出された類似性とセッションごとのログ情報の中から解析に必要な履歴情報とを用いて、ユーザに対する各コンテンツの推薦度を算出する（ステップＳ２０７）。 The history information collection unit 1201 of the short-term interest analysis processing unit 1210 collects history information necessary for analysis from the database 300 in real time from the log information for each session (step S204), and the similarity calculation unit 1202 collects the history information. The history information is analyzed to define the similarity between users (step S205). The recommendation level calculation unit 1203 calculates the recommendation level of each content for the user using the calculated similarity and history information necessary for analysis from the log information for each session (step S206). Also, the recommendation level calculation unit 1215 calculates the recommendation level of each content for the user using the similarity calculated by the similarity calculation unit 1102 and the history information necessary for analysis from the log information for each session. (Step S207).
　そして、レコメンド結果出力部１３１０は、統合割合算出部１６００の変更結果に応じて、推薦度算出部１２０３により算出したコンテンツごとの推薦度と推薦度算出部１２１５により算出したコンテンツごとの推薦度とからレコメンド結果を出力する（ステップＳ２０９）。 Then, the recommendation result output unit 1310 uses the recommendation level for each content calculated by the recommendation level calculation unit 1203 and the recommendation level for each content calculated by the recommendation level calculation unit 1215 according to the change result of the integration ratio calculation unit 1600. A recommendation result is output (step S209).
　以上、説明したように、本実施形態によれば、短期的興味分析処理部が、長期的嗜好分析処理で算出した類似性と短期的興味分析処理で算出した類似性とから推薦度を算出し、これらの推薦度とからレコメンド結果を出力するため、長期的嗜好分析処理部によって、ユーザの長期的な嗜好に基づくレコメンド結果を提供し、かつ、短期的興味分析処理部によってその状況においての短期的な興味に対するレコメンド結果を提供することにより、興味の遷移を考慮した適応的なレコメンド結果を提供することが可能となる。 As described above, according to the present embodiment, the short-term interest analysis processing unit calculates the recommendation degree from the similarity calculated by the long-term preference analysis processing and the similarity calculated by the short-term interest analysis processing. In order to output a recommendation result from these recommendation degrees, a long-term preference analysis processing unit provides a recommendation result based on the user's long-term preference, and a short-term interest analysis processing unit provides a short-term preference in the situation. By providing a recommendation result for a specific interest, it is possible to provide an adaptive recommendation result considering the transition of interest.
＜第３の実施形態＞
　図７および図９を用いて、本発明の第３の実施形態について説明する。 <Third Embodiment>
A third embodiment of the present invention will be described with reference to FIGS.
＜レコメンド装置の構成＞
　図７を用いて、本実施形態に係るレコメンド装置の構成について説明する。 <Configuration of recommendation device>
The configuration of the recommendation device according to the present embodiment will be described with reference to FIG.
　本実施形態に係るレコメンド装置１０００は、図７に示すように、長期的嗜好分析処理部１１２０と、短期的興味分析処理部１２２０と、レコメンド結果出力部１３２０と、レコメンド表示部１４００と、閲覧状況フィードバック部１５００と、統合割合算出部１６００とから構成され、長期的嗜好分析処理部１１２０には、データベース２００が、短期的興味分析処理部１２２０には、データベース３００が接続されている。 As shown in FIG. 7, the recommendation apparatus 1000 according to the present embodiment includes a long-term preference analysis processing unit 1120, a short-term interest analysis processing unit 1220, a recommendation result output unit 1320, a recommendation display unit 1400, and a browsing situation. The database 200 is connected to the long-term preference analysis processing unit 1120, and the database 300 is connected to the short-term interest analysis processing unit 1220.
＜レコメンド装置の詳細な構成＞
　本実施形態に係るレコメンド装置は、図８に示すように、長期的嗜好分析処理部１１２０を構成する履歴情報収集部１１０１と、類似性算出部１１０２と、推薦度算出部１１０３と、記憶部１１２６と、短期的興味分析処理部１２２０を構成する履歴情報収集部１２０１と、推薦度算出部１２２６と、レコメンド結果出力部１３２０と、レコメンド表示部１４００と、閲覧状況フィードバック部１５００と、統合割合算出部１６００とから構成されている。なお、第１の実施形態と同様の符号を付す構成要素については、同様の機能を有することから、その詳細な説明は、省略する。 <Detailed configuration of recommendation device>
As illustrated in FIG. 8, the recommendation device according to the present embodiment includes a history information collection unit 1101, a similarity calculation unit 1102, a recommendation degree calculation unit 1103, and a storage unit 1126 that form a long-term preference analysis processing unit 1120. A history information collection unit 1201, a recommendation degree calculation unit 1226, a recommendation result output unit 1320, a recommendation display unit 1400, a browsing status feedback unit 1500, and an integrated ratio calculation unit that constitute the short-term interest analysis processing unit 1220 1600. In addition, about the component which attaches | subjects the code | symbol similar to 1st Embodiment, since it has the same function, the detailed description is abbreviate | omitted.
　記憶部１１２６は、類似性算出部１１０２において定義した類似性および推薦度算出部１１０３において算出した推薦度を記憶する。推薦度算出部１２２６は、記憶部１１２６に記憶された類似性および算出した推薦度とセッションごとのログ情報の中から解析に必要な履歴情報とを用いて、ユーザに対する各コンテンツの推薦度を算出する。 The storage unit 1126 stores the similarity defined by the similarity calculation unit 1102 and the recommendation level calculated by the recommendation level calculation unit 1103. The recommendation level calculation unit 1226 calculates the recommendation level of each content for the user using the similarity stored in the storage unit 1126 and the calculated recommendation level and history information necessary for analysis from the log information for each session. To do.
　レコメンド結果出力部１３２０は、統合割合算出部１６００の変更結果に応じて、推薦度算出部１１０３により算出したコンテンツごとの推薦度と推薦度算出部１２２６により算出したコンテンツごとの推薦度とからレコメンド結果を出力する。 The recommendation result output unit 1320 determines a recommendation result from the recommendation level for each content calculated by the recommendation level calculation unit 1103 and the recommendation level for each content calculated by the recommendation level calculation unit 1226 according to the change result of the integration ratio calculation unit 1600. Is output.
＜レコメンド装置の処理＞
　図９を用いて、本実施形態に係るレコメンド装置の処理について説明する。 <Processing of recommendation device>
The process of the recommendation device according to the present embodiment will be described with reference to FIG.
　まず、長期的嗜好分析処理部１１２０の履歴情報収集部１１０１は、すべてのログ情報の中から解析に必要な履歴情報をデータベース２００から収集し（ステップＳ３０１）、類似性算出部１１０２は、収集した履歴情報を解析して、ユーザ間の類似性を定義する（ステップＳ３０２）。推薦度算出部１１０３は、算出された類似性とすべてのログ情報の中から解析に必要な履歴情報とを用いて、ユーザに対する各コンテンツの推薦度を算出し（ステップＳ３０３）、記憶部１１２６は、類似性算出部１１０２において定義した類似性および推薦度算出部１１０３において算出した推薦度を記憶する（ステップＳ３０４）。 First, the history information collection unit 1101 of the long-term preference analysis processing unit 1120 collects history information necessary for analysis from all the log information (step S301), and the similarity calculation unit 1102 collects the history information. The history information is analyzed to define the similarity between users (step S302). The recommendation level calculation unit 1103 calculates the recommendation level of each content for the user using the calculated similarity and history information necessary for analysis from all log information (step S303), and the storage unit 1126 The similarity defined in the similarity calculation unit 1102 and the recommendation level calculated in the recommendation level calculation unit 1103 are stored (step S304).
　短期的興味分析処理部１２２０の履歴情報収集部１２０１は、セッションごとのログ情報の中から解析に必要な履歴情報をデータベース３００からリアルタイムに収集する（ステップＳ３０５）。推薦度算出部１２２６は、記憶部１１２６に記憶された類似性および推薦度とセッションごとのログ情報の中から解析に必要な履歴情報とを用いて、ユーザに対する各コンテンツの推薦度を算出する（ステップＳ３０７）。 The history information collection unit 1201 of the short-term interest analysis processing unit 1220 collects history information necessary for analysis from the log information for each session from the database 300 in real time (step S305). The recommendation level calculation unit 1226 calculates the recommendation level of each content for the user using the similarity and recommendation level stored in the storage unit 1126 and the history information necessary for analysis from the log information for each session ( Step S307).
　そして、レコメンド結果出力部１３２０は、統合割合算出部１６００の変更結果に応じて、推薦度算出部１１０３により算出したコンテンツごとの推薦度と推薦度算出部１２２６により算出したコンテンツごとの推薦度とからレコメンド結果を出力する（ステップＳ３０８）。 Then, the recommendation result output unit 1320 uses the recommendation level for each content calculated by the recommendation level calculation unit 1103 and the recommendation level for each content calculated by the recommendation level calculation unit 1226 according to the change result of the integration ratio calculation unit 1600. A recommendation result is output (step S308).
　以上、説明したように、本実施形態によれば、長期的嗜好分析処理において算出した類似性を用いた推薦度と短期的興味分析処理が、長期的嗜好分析処理において算出した類似性を用いた推薦度とからレコメンド結果を出力するため、長期的嗜好分析処理部によって、ユーザの長期的な嗜好に基づくレコメンド結果を提供し、かつ、短期的興味分析処理部によってその状況においての短期的な興味に対するレコメンド結果を提供することにより、興味の遷移を考慮した適応的なレコメンド結果を提供することが可能となる。 As described above, according to the present embodiment, the recommendation degree using the similarity calculated in the long-term preference analysis process and the short-term interest analysis process use the similarity calculated in the long-term preference analysis process. In order to output the recommendation result based on the recommendation degree, the long-term preference analysis processing unit provides the recommendation result based on the long-term preference of the user, and the short-term interest analysis processing unit provides the short-term interest in the situation. By providing a recommendation result for, it is possible to provide an adaptive recommendation result considering the transition of interest.
＜第４の実施形態＞
　図１０から図１２を用いて、本発明の第４の実施形態について説明する。 <Fourth Embodiment>
A fourth embodiment of the present invention will be described with reference to FIGS.
＜レコメンドシステムの構成＞
　図１０を用いて、本実施形態に係るレコメンドシステムの構成について説明する。 <Configuration of recommendation system>
The configuration of the recommendation system according to the present embodiment will be described with reference to FIG.
　本実施形態に係るレコメンドシステムは、図１０に示すように、長期的嗜好分析処理装置４００と、短期的興味分析処理装置５００と、レコメンド結果出力装置６００と、レコメンド表示装置７００と、閲覧状況フィードバック装置８００と、統合割合算出装置９００とから構成され、長期的嗜好分析処理装置４００には、データベース２００が、短期的興味分析処理装置５００には、データベース３００が接続されている。 As shown in FIG. 10, the recommendation system according to the present embodiment includes a long-term preference analysis processing device 400, a short-term interest analysis processing device 500, a recommendation result output device 600, a recommendation display device 700, and browsing status feedback. The long-term preference analysis processing device 400 is connected to the database 200, and the short-term interest analysis processing device 500 is connected to the database 300.
　なお、本実施形態に係るレコメンドシステムは、レコメンド処理を長期的嗜好分析処理装置と短期的興味分析処理装置とに分け、長期的嗜好分析処理装置で、長期的な嗜好分析処理を行うとともに、短期的興味分析処理装置で、短期的な嗜好分析処理を行い、長期的嗜好分析処理装置４００の結果と短期的興味分析処理装置５００の結果とをレコメンド結果出力装置６００で統合して出力する。なお、長期的嗜好分析処理装置４００としては、演算処理能力が高いクラウドサーバ、短期的興味分析処理装置５００としては、長期的嗜好分析処理装置４００同様に演算処理能力が高いクラウドサーバに加えて、比較的演算処理能力が低い携帯端末やＳＴＢ（Ｓｅｔ　Ｔｏｐ　Ｂｏｘ）等での実施が例示できるが、これに限るものではない。 The recommendation system according to the present embodiment divides the recommendation processing into a long-term preference analysis processing device and a short-term interest analysis processing device, and performs long-term preference analysis processing with the long-term preference analysis processing device. A short-term preference analysis process is performed by the automatic interest analysis processing device, and the result of the long-term preference analysis processing device 400 and the result of the short-term interest analysis processing device 500 are integrated by the recommendation result output device 600 and output. In addition, as a long-term preference analysis processing device 400, in addition to a cloud server with high calculation processing capability, and as a short-term interest analysis processing device 500, in addition to a cloud server with high calculation processing capability like the long-term preference analysis processing device 400, Although implementation with a portable terminal or STB (Set Top Box) having relatively low arithmetic processing capability can be exemplified, the present invention is not limited to this.
＜レコメンドシステムの詳細な構成＞
　本実施形態に係るレコメンドシステムは、図１１に示すように、長期的嗜好分析処理装置４００を構成する履歴情報収集部４０１と、類似性算出部４０２と、推薦度算出部４０３と、記憶部４０４と、短期的興味分析処理装置５００を構成する履歴情報収集部５０１と、類似性算出部５０２と、推薦度算出部５０３と、レコメンド結果出力装置６００と、レコメンド表示装置７００と、閲覧状況フィードバック装置８００と、統合割合算出装置９００とから構成されている。 <Detailed configuration of recommendation system>
As shown in FIG. 11, the recommendation system according to the present embodiment includes a history information collection unit 401, a similarity calculation unit 402, a recommendation degree calculation unit 403, and a storage unit 404 that constitute the long-term preference analysis processing device 400. A history information collection unit 501, a similarity calculation unit 502, a recommendation degree calculation unit 503, a recommendation result output device 600, a recommendation display device 700, and a browsing status feedback device that constitute the short-term interest analysis processing device 500. 800 and an integrated ratio calculation device 900.
　履歴情報収集部４０１は、すべてのログ情報の中から解析に必要な履歴情報をデータベース２００から収集する。類似性算出部４０２は、収集した履歴情報を解析して、ユーザ間の類似性を定義する。推薦度算出部４０３は、算出された類似性とすべてのログ情報の中から収集された解析に必要な履歴情報とを用いて、ユーザに対する各コンテンツの推薦度を算出する。記憶部４０４は、算出した推薦度を記憶する。 The history information collection unit 401 collects history information necessary for analysis from all the log information from the database 200. The similarity calculation unit 402 analyzes the collected history information and defines the similarity between users. The recommendation level calculation unit 403 calculates the recommendation level of each content for the user using the calculated similarity and history information necessary for analysis collected from all log information. The storage unit 404 stores the calculated recommendation level.
　履歴情報収集部５０１は、セッションごとのログ情報の中から解析に必要な履歴情報をデータベース３００リアルタイムに収集する。類似性算出部５０２は、収集した履歴情報を解析して、ユーザ間の類似性を定義する。推薦度算出部５０３は、算出された類似性とセッションごとのログ情報の中から解析に必要な履歴情報とを用いて、ユーザに対する各コンテンツの推薦度を算出する。 The history information collection unit 501 collects history information necessary for analysis from the log information for each session in real time in the database 300. The similarity calculation unit 502 analyzes the collected history information and defines the similarity between users. The recommendation level calculation unit 503 calculates the recommendation level of each content for the user using the calculated similarity and history information necessary for analysis from the log information for each session.
　レコメンド結果出力装置６００は、統合割合算出装置９００の変更結果に応じて、記憶されたコンテンツごとの推薦度と推薦度算出部５０３により算出したコンテンツごとの推薦度とからレコメンド結果を出力する。なお、統合割合算出装置９００の機能は、加重和を変化させることであり、例えば、セッションの閲覧個数やセッションタイム等に基づいて、指数・ロジスティック関数等により処理される。 The recommendation result output device 600 outputs a recommendation result from the stored recommendation level for each content and the recommendation level for each content calculated by the recommendation level calculation unit 503 according to the change result of the integrated ratio calculation device 900. Note that the function of the integrated ratio calculation apparatus 900 is to change the weighted sum, and is processed by an index / logistic function or the like based on, for example, the number of sessions viewed or the session time.
[規則26に基づく補充 10.01.2013]　
＜レコメンドシステムの処理＞
　図１２を用いて、本実施形態に係るレコメンドシステムの処理について説明する。[Supplement under rule 26 10.01.2013]
<Recommendation system processing>
Processing of the recommendation system according to the present embodiment will be described with reference to FIG.
　まず、長期的嗜好分析処理装置４００の履歴情報収集部４０１は、すべてのログ情報の中から解析に必要な履歴情報をデータベース２００から収集し（ステップＳ４０１）、類似性算出部４０２は、収集した履歴情報を解析して、ユーザ間の類似性を、例えば、二人のユーザの履歴を比較して同時に履歴を持つコンテンツ数を元に、または、二つのコンテンツの履歴を比較して同時に履歴を持つコンテンツ数を元に定義する（ステップＳ４０２）。推薦度算出部４０３は、算出された類似性とすべてのログ情報の中から解析に必要な履歴情報とを用いて、ユーザに対する各コンテンツの推薦度を、例えば、当該ユーザのコンテンツの履歴を持つ他のユーザとの類似性、または、当該ユーザの履歴に存在するコンテンツと他のコンテンツとの類似性を元に算出し（ステップＳ４０３）、記憶部４０４は、算出した推薦度を記憶する（ステップＳ４０４）。 First, the history information collection unit 401 of the long-term preference analysis processing device 400 collects history information necessary for analysis from all log information from the database 200 (step S401), and the similarity calculation unit 402 collects the history information. Analyze history information and compare similarities between users, for example, compare the history of two users based on the number of content with history at the same time, or compare the history of two content and The definition is made based on the number of contents possessed (step S402). The recommendation level calculation unit 403 uses the calculated similarity and history information necessary for analysis from all log information, and has a recommendation level of each content for the user, for example, a history of the content of the user. Calculation is performed based on the similarity with other users or the similarity between the content existing in the user's history and other content (step S403), and the storage unit 404 stores the calculated recommendation degree (step S403). S404).
　短期的興味分析処理装置５００の履歴情報収集部５０１は、セッションごとのログ情報の中から解析に必要な履歴情報をデータベース３００から収集し（ステップＳ４０５）、類似性算出部５０２は、収集した履歴情報を解析して、ユーザ間の類似性を、前述同様に定義する（ステップＳ４０６）。推薦度算出部５０３は、算出された類似性とセッションごとのログ情報の中から解析に必要な履歴情報とを用いて、ユーザに対する各コンテンツの推薦度を算出する（ステップＳ４０７）。 The history information collection unit 501 of the short-term interest analysis processing apparatus 500 collects history information necessary for analysis from the log information for each session (step S405), and the similarity calculation unit 502 collects the collected history information. By analyzing the information, the similarity between the users is defined as described above (step S406). The recommendation level calculation unit 503 calculates the recommendation level of each content for the user using the calculated similarity and history information necessary for analysis from the log information for each session (step S407).
　そして、レコメンド結果出力装置６００は、統合割合算出装置９００の変更結果に応じて、記憶されたコンテンツごとの推薦度と推薦度算出部５０３により算出したコンテンツごとの推薦度とからレコメンド結果を出力する（ステップＳ４０８）。 Then, the recommendation result output device 600 outputs a recommendation result from the stored recommendation level for each content and the recommendation level for each content calculated by the recommendation level calculation unit 503 according to the change result of the integration ratio calculation device 900. (Step S408).
　以上、説明したように、本実施形態によれば、長期的嗜好分析処理装置において算出した類似性を用いた推薦度と、短期的興味分析処理装置が長期的嗜好分析処理装置において算出した類似性を用いた推薦度とからレコメンド結果を出力するため、長期的嗜好分析処理装置によって、ユーザの長期的な嗜好に基づくレコメンド結果を提供し、かつ、短期的興味分析処理装置によってその状況においての短期的な興味に対するレコメンド結果を提供することにより、興味の遷移を考慮した適応的なレコメンド結果を提供することが可能となる。 As described above, according to the present embodiment, the degree of recommendation using the similarity calculated in the long-term preference analysis processing device and the similarity calculated by the short-term interest analysis processing device in the long-term preference analysis processing device. In order to output recommendation results based on the degree of recommendation using, the long-term preference analysis processing device provides recommendation results based on the user's long-term preference, and the short-term interest analysis processing device provides a short-term in that situation. By providing a recommendation result for a specific interest, it is possible to provide an adaptive recommendation result considering the transition of interest.
＜第５の実施形態＞
　図１３から図１５を用いて、本発明の第５の実施形態について説明する。 <Fifth Embodiment>
A fifth embodiment of the present invention will be described with reference to FIGS.
＜レコメンドシステムの構成＞
　図１３を用いて、本実施形態に係るレコメンドシステムの構成について説明する。 <Configuration of recommendation system>
The configuration of the recommendation system according to the present embodiment will be described with reference to FIG.
　本実施形態に係るレコメンドシステムは、図１３に示すように、長期的嗜好分析処理装置４１０と、短期的興味分析処理装置５１０と、レコメンド結果出力装置６１０と、レコメンド表示装置７００と、閲覧状況フィードバック装置８００と、統合割合算出装置９００とから構成され、長期的嗜好分析処理装置４１０には、データベース２００が、短期的興味分析処理装置５１０には、データベース３００が接続されている。 As illustrated in FIG. 13, the recommendation system according to the present embodiment includes a long-term preference analysis processing device 410, a short-term interest analysis processing device 510, a recommendation result output device 610, a recommendation display device 700, and browsing status feedback. The long-term preference analysis processing device 410 is connected to the database 200, and the short-term interest analysis processing device 510 is connected to the database 300.
＜レコメンドシステムの詳細な構成＞
　本実施形態に係るレコメンドシステムは、図１４に示すように、長期的嗜好分析処理装置４１０を構成する履歴情報収集部４０１と、類似性算出部４０２と、記憶部４１５と、短期的興味分析処理装置５１０を構成する履歴情報収集部５０１と、類似性算出部５０２と、推薦度算出部５０３と、取得部５１４と、推薦度算出部５１５と、レコメンド結果出力装置６１０と、レコメンド表示装置７００と、閲覧状況フィードバック装置８００と、統合割合算出装置９００とから構成されている。なお、第４の実施形態と同様の符号を付す構成要素については、同様の機能を有することから、その詳細な説明は、省略する。 <Detailed configuration of recommendation system>
As shown in FIG. 14, the recommendation system according to the present embodiment includes a history information collection unit 401, a similarity calculation unit 402, a storage unit 415, and a short-term interest analysis process that constitute the long-term preference analysis processing device 410. A history information collection unit 501, a similarity calculation unit 502, a recommendation degree calculation unit 503, an acquisition unit 514, a recommendation degree calculation unit 515, a recommendation result output device 610, and a recommendation display device 700 that constitute the device 510. The browsing status feedback device 800 and the integrated ratio calculation device 900 are configured. In addition, about the component which attaches | subjects the code | symbol similar to 4th Embodiment, since it has the same function, the detailed description is abbreviate | omitted.
　記憶部４１５は、類似性算出部４０２が算出した類似性を記憶する。推薦度算出部５１５は、記憶部４１５に記憶されている類似性と、セッションごとのログ情報の中から解析に必要な履歴情報とを用いて、ユーザに対する各コンテンツの推薦度を算出する。 The storage unit 415 stores the similarity calculated by the similarity calculation unit 402. The recommendation level calculation unit 515 calculates the recommendation level of each content for the user using the similarity stored in the storage unit 415 and the history information necessary for analysis from the log information for each session.
　レコメンド結果出力装置６１０は、統合割合算出装置９００の変更結果に応じて、推薦度算出部５０３により算出したコンテンツごとの推薦度と推薦度算出部５１５により算出したコンテンツごとの推薦度とからレコメンド結果を出力する。 The recommendation result output device 610 determines a recommendation result from the recommendation level for each content calculated by the recommendation level calculation unit 503 and the recommendation level for each content calculated by the recommendation level calculation unit 515 according to the change result of the integration ratio calculation device 900. Is output.
[規則26に基づく補充 10.01.2013]　
＜レコメンドシステムの処理＞
　図１５を用いて、本実施形態に係るレコメンドシステムの処理について説明する。[Supplement under rule 26 10.01.2013]
<Recommendation system processing>
Processing of the recommendation system according to the present embodiment will be described with reference to FIG.
　まず、長期的嗜好分析処理装置４１０の履歴情報収集部４０１は、すべてのログ情報の中から解析に必要な履歴情報をデータベース２００から収集し（ステップＳ５０１）、類似性算出部４０２は、収集した履歴情報を解析して、ユーザ間の類似性を定義し（ステップＳ５０２）、記憶部４１５は、算出した類似性を記憶する（ステップＳ５０３）。 First, the history information collection unit 401 of the long-term preference analysis processing apparatus 410 collects history information necessary for analysis from all log information from the database 200 (step S501), and the similarity calculation unit 402 collects the history information. The history information is analyzed to define the similarity between users (step S502), and the storage unit 415 stores the calculated similarity (step S503).
　短期的興味分析処理装置５１０の履歴情報収集部５０１は、セッションごとのログ情報の中から解析に必要な履歴情報をデータベース３００からリアルタイムに収集し（ステップＳ５０４）、類似性算出部５０２は、収集した履歴情報を解析して、ユーザ間の類似性を定義する（ステップＳ５０５）。推薦度算出部５０３は、算出された類似性とセッションごとのログ情報の中から解析に必要な履歴情報とを用いて、ユーザに対する各コンテンツの推薦度を算出する（ステップＳ５０６）。また、推薦度算出部５１５は、類似性算出部４０２により算出された類似性とセッションごとのログ情報の中から解析に必要な履歴情報とを用いて、ユーザに対する各コンテンツの推薦度を算出する（ステップＳ５０７）。 The history information collection unit 501 of the short-term interest analysis processing apparatus 510 collects history information necessary for analysis from the log information for each session in real time from the database 300 (step S504), and the similarity calculation unit 502 collects the history information. The history information is analyzed to define similarity between users (step S505). The recommendation level calculation unit 503 calculates the recommendation level of each content for the user using the calculated similarity and history information necessary for analysis from the log information for each session (step S506). Also, the recommendation level calculation unit 515 calculates the recommendation level of each content for the user using the similarity calculated by the similarity calculation unit 402 and the history information necessary for analysis from the log information for each session. (Step S507).
　そして、レコメンド結果出力装置６１０は、統合割合算出装置９００の変更結果に応じて、推薦度算出部５０３により算出したコンテンツごとの推薦度と推薦度算出部５１５により算出したコンテンツごとの推薦度とからレコメンド結果を出力する（ステップＳ５０９）。 Then, the recommendation result output device 610 uses the recommendation level for each content calculated by the recommendation level calculation unit 503 and the recommendation level for each content calculated by the recommendation level calculation unit 515 according to the change result of the integrated ratio calculation device 900. A recommendation result is output (step S509).
　以上、説明したように、本実施形態によれば、長期的嗜好分析処理装置において算出した類似性を用いた推薦度と、短期的興味分析処理装置が長期的嗜好分析処理装置において算出した類似性を用いた推薦度とからレコメンド結果を出力するため、長期的嗜好分析処理装置によって、ユーザの長期的な嗜好に基づくレコメンド結果を提供し、かつ、短期的興味分析処理装置によってその状況においての短期的な興味に対するレコメンド結果を提供することにより、興味の遷移を考慮した適応的なレコメンド結果を提供することが可能となる。 As described above, according to the present embodiment, the degree of recommendation using the similarity calculated in the long-term preference analysis processing device and the similarity calculated by the short-term interest analysis processing device in the long-term preference analysis processing device. In order to output recommendation results based on the degree of recommendation using, the long-term preference analysis processing device provides recommendation results based on the user's long-term preference, and the short-term interest analysis processing device provides a short-term in that situation. By providing a recommendation result for a specific interest, it is possible to provide an adaptive recommendation result considering the transition of interest.
＜第６の実施形態＞
　図１６から図１８を用いて、本発明の第６の実施形態について説明する。 <Sixth Embodiment>
A sixth embodiment of the present invention will be described with reference to FIGS.
＜レコメンドシステムの構成＞
　図１６を用いて、本実施形態に係るレコメンドシステムの構成について説明する。 <Configuration of recommendation system>
The configuration of the recommendation system according to this embodiment will be described with reference to FIG.
　本実施形態に係るレコメンドシステムは、図１６に示すように、長期的嗜好分析処理装置４２０と、短期的興味分析処理装置５２０と、レコメンド結果出力装置６２０と、レコメンド表示装置７００と、閲覧状況フィードバック装置８００と、統合割合算出装置９００とから構成され、長期的嗜好分析処理装置４２０には、データベース２００が、短期的興味分析処理装置５２０には、データベース３００が接続されている。 As illustrated in FIG. 16, the recommendation system according to the present embodiment includes a long-term preference analysis processing device 420, a short-term interest analysis processing device 520, a recommendation result output device 620, a recommendation display device 700, and browsing status feedback. The long-term preference analysis processing device 420 is connected to the database 200, and the short-term interest analysis processing device 520 is connected to the database 300.
＜レコメンドシステムの詳細な構成＞
　本実施形態に係るレコメンドシステムは、図１７に示すように、長期的嗜好分析処理装置４２０を構成する履歴情報収集部４０１と、類似性算出部４０２と、推薦度算出部４０３と、記憶部４２６と、短期的興味分析処理装置５２０を構成する履歴情報収集部５０１と、推薦度算出部５２６と、取得部５２７と、レコメンド結果出力装置６２０と、レコメンド表示装置７００と、閲覧状況フィードバック装置８００と、統合割合算出装置９００とから構成されている。なお、第４の実施形態と同様の符号を付す構成要素については、同様の機能を有することから、その詳細な説明は、省略する。 <Detailed configuration of recommendation system>
As illustrated in FIG. 17, the recommendation system according to the present embodiment includes a history information collection unit 401, a similarity calculation unit 402, a recommendation degree calculation unit 403, and a storage unit 426 that constitute the long-term preference analysis processing device 420. A history information collection unit 501, a recommendation degree calculation unit 526, an acquisition unit 527, a recommendation result output device 620, a recommendation display device 700, and a browsing status feedback device 800 that constitute the short-term interest analysis processing device 520. , And an integrated ratio calculation device 900. In addition, about the component which attaches | subjects the code | symbol similar to 4th Embodiment, since it has the same function, the detailed description is abbreviate | omitted.
　記憶部４２６は、類似性算出部４０２において定義した類似性および推薦度算出部４０３において算出した推薦度を記憶する。推薦度算出部５２６は、記憶部４２６に記憶された類似性および算出した推薦度とセッションごとのログ情報の中から解析に必要な履歴情報とを用いて、ユーザに対する各コンテンツの推薦度を算出する。 The storage unit 426 stores the similarity defined by the similarity calculation unit 402 and the recommendation level calculated by the recommendation level calculation unit 403. The recommendation level calculation unit 526 calculates the recommendation level of each content for the user using the similarity stored in the storage unit 426 and the calculated recommendation level and history information necessary for analysis from the log information for each session. To do.
　レコメンド結果出力装置６２０は、統合割合算出装置９００の変更結果に応じて、推薦度算出部４０３により算出したコンテンツごとの推薦度と推薦度算出部５２６により算出したコンテンツごとの推薦度とからレコメンド結果を出力する。 The recommendation result output device 620 determines a recommendation result from the recommendation level for each content calculated by the recommendation level calculation unit 403 and the recommendation level for each content calculated by the recommendation level calculation unit 526 according to the change result of the integration ratio calculation device 900. Is output.
＜レコメンドシステムの処理＞
　図１８を用いて、本実施形態に係るレコメンドシステムの処理について説明する。 <Recommendation system processing>
Processing of the recommendation system according to the present embodiment will be described with reference to FIG.
　まず、長期的嗜好分析処理装置４２０の履歴情報収集部４０１は、すべてのログ情報の中から解析に必要な履歴情報をデータベース２００から収集し（ステップＳ６０１）、類似性算出部４０２は、収集した履歴情報を解析して、ユーザ間の類似性を定義する（ステップＳ６０２）。推薦度算出部４０３は、算出された類似性とすべてのログ情報の中から解析に必要な履歴情報とを用いて、ユーザに対する各コンテンツの推薦度を算出し（ステップＳ６０３）、記憶部４２６は、類似性算出部４０２において定義した類似性および推薦度算出部４０３において算出した推薦度を記憶する（ステップＳ６０４）。 First, the history information collection unit 401 of the long-term preference analysis processing device 420 collects history information necessary for analysis from all log information from the database 200 (step S601), and the similarity calculation unit 402 collects the history information. The history information is analyzed to define the similarity between users (step S602). The recommendation level calculation unit 403 calculates the recommendation level of each content for the user using the calculated similarity and history information necessary for analysis from all the log information (step S603), and the storage unit 426 The similarity defined in the similarity calculation unit 402 and the recommendation level calculated in the recommendation level calculation unit 403 are stored (step S604).
　短期的興味分析処理装置５２０の履歴情報収集部５０１は、セッションごとのログ情報の中から解析に必要な履歴情報をデータベース３００からリアルタイムに収集する（ステップＳ６０５）。推薦度算出部５２６は、記憶部４２６に記憶された類似性および推薦度とセッションごとのログ情報の中から解析に必要な履歴情報とを用いて、ユーザに対する各コンテンツの推薦度を算出する（ステップＳ６０７）。 The history information collection unit 501 of the short-term interest analysis processing device 520 collects history information necessary for analysis from the database 300 in real time from the log information for each session (step S605). The recommendation level calculation unit 526 calculates the recommendation level of each content for the user using the similarity and recommendation level stored in the storage unit 426 and the history information necessary for analysis from the log information for each session ( Step S607).
　そして、レコメンド結果出力装置６２０は、統合割合算出装置９００の変更結果に応じて、推薦度算出部４０３により算出したコンテンツごとの推薦度と推薦度算出部５２６により算出したコンテンツごとの推薦度とからレコメンド結果を出力する（ステップＳ６０８）。 Then, the recommendation result output device 620 determines from the recommendation level for each content calculated by the recommendation level calculation unit 403 and the recommendation level for each content calculated by the recommendation level calculation unit 526 according to the change result of the integrated ratio calculation device 900. A recommendation result is output (step S608).
　以上、説明したように、本実施形態によれば、長期的嗜好分析処理装置において算出した類似性を用いた推薦度と、短期的興味分析処理装置が長期的嗜好分析処理装置において算出した類似性を用いた推薦度とからレコメンド結果を出力するため、長期的嗜好分析処理装置によって、ユーザの長期的な嗜好に基づくレコメンド結果を提供し、かつ、短期的興味分析処理装置によってその状況においての短期的な興味に対するレコメンド結果を提供することにより、興味の遷移を考慮した適応的なレコメンド結果を提供することが可能となる。 As described above, according to the present embodiment, the degree of recommendation using the similarity calculated in the long-term preference analysis processing device and the similarity calculated by the short-term interest analysis processing device in the long-term preference analysis processing device. In order to output recommendation results based on the degree of recommendation using, the long-term preference analysis processing device provides recommendation results based on the user's long-term preference, and the short-term interest analysis processing device provides a short-term in that situation. By providing a recommendation result for a specific interest, it is possible to provide an adaptive recommendation result considering the transition of interest.
　なお、レコメンド装置あるいはレコメンドシステムの処理をコンピュータ読み取り可能な記録媒体に記録し、この記録媒体に記録されたプログラムをレコメンド装置あるいは長期的嗜好分析処理装置、短期的興味分析処理装置、レコメンド結果出力装置等に読み込ませ、実行することによって本発明のレコメンド装置あるいはレコメンドシステムを実現することができる。ここでいうコンピュータシステムとは、ＯＳや周辺装置等のハードウェアを含む。 The processing of the recommendation device or the recommendation system is recorded on a computer-readable recording medium, and the program recorded on the recording medium is recommended, the long-term preference analysis processing device, the short-term interest analysis processing device, or the recommendation result output device. The recommendation device or the recommendation system of the present invention can be realized by reading and executing the above. The computer system here includes an OS and hardware such as peripheral devices.
　また、「コンピュータシステム」は、ＷＷＷ（Ｗｏｒｌｄ　Ｗｉｄｅ　Ｗｅｂ）システムを利用している場合であれば、ホームページ提供環境（あるいは表示環境）も含むものとする。また、上記プログラムは、このプログラムを記憶装置等に格納したコンピュータシステムから、伝送媒体を介して、あるいは、伝送媒体中の伝送波により他のコンピュータシステムに伝送されても良い。ここで、プログラムを伝送する「伝送媒体」は、インターネット等のネットワーク（通信網）や電話回線等の通信回線（通信線）のように情報を伝送する機能を有する媒体のことをいう。 In addition, the “computer system” includes a homepage providing environment (or display environment) if a WWW (World Wide Web) system is used. The program may be transmitted from a computer system storing the program in a storage device or the like to another computer system via a transmission medium or by a transmission wave in the transmission medium. Here, the “transmission medium” for transmitting the program refers to a medium having a function of transmitting information, such as a network (communication network) such as the Internet or a communication line (communication line) such as a telephone line.
　また、上記プログラムは、前述した機能の一部を実現するためのものであっても良い。さらに、前述した機能をコンピュータシステムにすでに記録されているプログラムとの組合せで実現できるもの、いわゆる差分ファイル（差分プログラム）であっても良い。 Further, the program may be for realizing a part of the above-described functions. Furthermore, what can implement | achieve the function mentioned above in combination with the program already recorded on the computer system, and what is called a difference file (difference program) may be sufficient.
　以上、この発明の実施形態につき、図面を参照して詳述してきたが、具体的な構成はこの実施形態に限られるものではなく、この発明の要旨を逸脱しない範囲の設計等も含まれる。例えば、本発明では、利用する履歴情報はデータベース等の記憶媒体で集約管理されており、履歴情報収集部を用いてこれらを取得することを想定しているが、必ずしもこれに限るものではない。 As described above, the embodiments of the present invention have been described in detail with reference to the drawings. However, the specific configuration is not limited to the embodiments, and includes designs and the like that do not depart from the gist of the present invention. For example, in the present invention, history information to be used is centrally managed in a storage medium such as a database, and it is assumed that these are acquired using a history information collection unit. However, the present invention is not necessarily limited to this.
　また、本発明によるレコメンド結果の提供方法は、当該ユーザに対する一つ一つのコンテンツの推薦度を記したリストを、外部のレコメンド提示手法へと提供することを想定しているが、必ずしもこの利用方法に限るものではない。 Further, the method for providing recommendation results according to the present invention assumes that a list describing the degree of recommendation of each content for the user is provided to an external recommendation presentation method. It is not limited to.
　なお、本発明のレコメンドの各処理は、一サーバによる実施、又は、機能別に特化した複数サーバの連携、若しくは負荷分散機能を用いた複数サーバの連携による実施等の実施方法であってもよい。 It should be noted that each process of the recommendation according to the present invention may be performed by one server, or by an implementation method such as implementation by cooperation of a plurality of servers specialized for each function or cooperation of a plurality of servers using a load distribution function. .
　１０００：レコメンド装置
　１１００：長期的嗜好分析処理部
　１１１０：長期的嗜好分析処理部
　１１２０：長期的嗜好分析処理部
　１１０１：履歴情報収集部
　１１０２：類似性算出部
　１１０３：推薦度算出部
　１１０４：記憶部
　１１１５：記憶部
　１１２６：記憶部
　１２００：短期的興味分析処理部
　１２１０：短期的興味分析処理部
　１２２０：短期的興味分析処理部
　１２０１：履歴情報収集部
　１２０２：類似性算出部
　１２０３：推薦度算出部
　１２１５：推薦度算出部
　１２２６：推薦度算出部
　１３００：レコメンド結果出力部
　１３１０：レコメンド結果出力部
　１３２０：レコメンド結果出力部
　１４００：レコメンド表示部
　１５００：閲覧状況フィードバック部
　１６００：統合割合算出部
　２００：データベース
　３００：データベース
　４００：長期的嗜好分析処理装置
　４１０：長期的嗜好分析処理装置
　４２０：長期的嗜好分析処理装置
　４０１：履歴情報収集部
　４０２：類似性算出部
　４０３：推薦度算出部
　４０４：記憶部
　４１５：記憶部
　４２６：記憶部
　５００：短期的興味分析処理装置
　５１０：短期的興味分析処理装置
　５２０：短期的興味分析処理装置
　５０１：履歴情報収集部
　５０２：類似性算出部
　５０３：推薦度算出部
　５１５：推薦度算出部
　５２６：推薦度算出部
　６００：レコメンド結果出力装置
　６１０：レコメンド結果出力装置
　６２０：レコメンド結果出力装置
　７００：レコメンド表示装置
　８００：閲覧状況フィードバック装置
　９００：統合割合算出装置 1000: recommendation device 1100: long-term preference analysis processing unit 1110: long-term preference analysis processing unit 1120: long-term preference analysis processing unit 1101: history information collection unit 1102: similarity calculation unit 1103: recommendation degree calculation unit 1104: storage unit 1115: Storage unit 1126: Storage unit 1200: Short-term interest analysis processing unit 1210: Short-term interest analysis processing unit 1220: Short-term interest analysis processing unit 1201: History information collection unit 1202: Similarity calculation unit 1203: Recommendation degree calculation unit 1215: Recommendation degree calculation unit 1226: Recommendation degree calculation unit 1300: Recommendation result output unit 1310: Recommendation result output unit 1320: Recommendation result output unit 1400: Recommendation display unit 1500: Browsing status feedback unit 1600: Integration ratio calculation unit 200: Database 300: Database 4 0: Long-term preference analysis processing device 410: Long-term preference analysis processing device 420: Long-term preference analysis processing device 401: History information collection unit 402: Similarity calculation unit 403: Recommendation degree calculation unit 404: Storage unit 415: Storage unit 426: storage unit 500: short-term interest analysis processing device 510: short-term interest analysis processing device 520: short-term interest analysis processing device 501: history information collection unit 502: similarity calculation unit 503: recommendation degree calculation unit 515: recommendation degree Calculation unit 526: recommendation degree calculation unit 600: recommendation result output device 610: recommendation result output device 620: recommendation result output device 700: recommendation display device 800: browsing status feedback device 900: integrated ratio calculation device

Claims (12)

  1. 　購買履歴や閲覧履歴等のユーザの履歴情報を元にユーザ間の類似性を定義し、該定義した類似性に基づいてユーザが興味を持つと予測されるコンテンツをレコメンド結果として出力する協調フィルタリング方法を用いるレコメンド装置であって、
    　すべての前記履歴情報を解析してレコメンドに必要な情報を生成する長期的嗜好分析処理手段と、
    　セッションごとの履歴情報を解析してレコメンドに必要な情報を生成する短期的興味分析処理手段と、
    　前記長期的嗜好分析処理手段の出力または／および前記短期的興味分析処理手段の出力を統合してレコメンド結果を出力するレコメンド結果出力手段と、
    　前記レコメンド結果出力手段のレコメンド結果をユーザの興味の遷移に基づいて変更する統合割合算出手段と、
    　を備えたことを特徴とするレコメンド装置。     A collaborative filtering method that defines similarity between users based on user history information such as purchase history and browsing history, and outputs content predicted to be of interest to the user based on the defined similarity as a recommendation result A recommendation device using
    Long-term preference analysis processing means for analyzing all the history information and generating information necessary for recommendation;
    A short-term interest analysis processing means that analyzes history information for each session and generates information necessary for the recommendation,
    A recommendation result output means for outputting a recommendation result by integrating the output of the long-term preference analysis processing means and / or the output of the short-term interest analysis processing means;
    An integrated ratio calculating means for changing the recommendation result of the recommendation result output means based on a transition of the user's interest;
    A recommendation device characterized by comprising:
  2. 　前記長期的嗜好分析処理手段が、
    　すべてのログ情報の中から解析に必要な履歴情報を収集する第１の履歴情報収集手段と、
    　該収集した履歴情報を解析して、ユーザ間の類似性を定義する第１の類似性算出手段と、
    　該算出された類似性と該すべてのログ情報の中から収集された解析に必要な履歴情報とを用いて、ユーザに対する各コンテンツの推薦度を算出する第１の推薦度算出手段と、
    　該算出した推薦度を記憶する記憶手段と、
    　を備え、
    　前記短期的興味分析処理手段が、
    　セッションごとのログ情報の中から解析に必要な履歴情報をリアルタイムに収集する第２の履歴情報収集手段と、
    　該収集した履歴情報を解析して、ユーザ間の類似性を定義する第２の類似性算出手段と、
    　該算出された類似性と該セッションごとのログ情報の中から解析に必要な履歴情報とを用いて、ユーザに対する各コンテンツの推薦度を算出する第２の推薦度算出手段と、
    　を備え、
    　前記レコメンド結果出力手段が、前記統合割合算出手段の変更結果に応じて、前記記憶されたコンテンツごとの推薦度と前記第２の推薦度算出手段により算出したコンテンツごとの推薦度とからレコメンド結果を出力することを特徴とする請求項１に記載のレコメンド装置。     The long-term preference analysis processing means is
    A first history information collecting means for collecting history information necessary for analysis from all log information;
    Analyzing the collected history information and defining first similarity between users; and
    First recommendation degree calculation means for calculating a recommendation degree of each content for the user using the calculated similarity and history information necessary for analysis collected from all the log information;
    Storage means for storing the calculated recommendation degree;
    With
    The short-term interest analysis processing means includes:
    A second history information collecting means for collecting history information necessary for analysis from log information for each session in real time;
    A second similarity calculation means for analyzing the collected history information and defining the similarity between users;
    Second recommendation degree calculating means for calculating a recommendation degree of each content for the user using the calculated similarity and history information necessary for analysis from the log information for each session;
    With
    The recommendation result output means outputs a recommendation result from the stored recommendation degree for each content and the recommendation degree for each content calculated by the second recommendation degree calculation means according to the change result of the integration ratio calculation means. The recommendation device according to claim 1, wherein the recommendation device outputs the recommendation device.
  3. 　前記長期的嗜好分析処理手段が、
    　すべてのログ情報の中から解析に必要な履歴情報を収集する第１の履歴情報収集手段と、
    　該収集した履歴情報を解析して、ユーザ間の類似性を定義する第１の類似性算出手段と、
    　該算出した類似性を記憶する記憶手段と、
    　を備え、
    　前記短期的興味分析処理手段が、
    　セッションごとのログ情報の中から解析に必要な履歴情報をリアルタイムに収集する第２の履歴情報収集手段と、
    　該収集した履歴情報を解析して、ユーザ間の類似性を定義する第２の類似性算出手段と、
    　該セッションごとのログ情報の中から解析に必要な履歴情報と前記記憶された類似性とを用いて、ユーザに対する各コンテンツの推薦度を算出する第１の推薦度算出手段と、
    　前記第２の類似性算出手段により算出された類似性と前記セッションごとのログ情報の中から解析に必要な履歴情報とを用いて、ユーザに対する各コンテンツの推薦度を算出する第２の推薦度算出手段と、
    　を備え、
    　前記レコメンド結果出力手段が、前記統合割合算出手段の変更結果に応じて、前記第１の推薦度算出手段により算出したコンテンツごとの推薦度と前記第２の推薦度算出手段により算出したコンテンツごとの推薦度とからレコメンド結果を出力することを特徴とする請求項１に記載のレコメンド装置。     The long-term preference analysis processing means is
    A first history information collecting means for collecting history information necessary for analysis from all log information;
    Analyzing the collected history information and defining first similarity between users; and
    Storage means for storing the calculated similarity;
    With
    The short-term interest analysis processing means includes:
    A second history information collecting means for collecting history information necessary for analysis from log information for each session in real time;
    A second similarity calculation means for analyzing the collected history information and defining the similarity between users;
    First recommendation degree calculating means for calculating a recommendation degree of each content for a user using history information necessary for analysis from the log information for each session and the stored similarity;
    A second recommendation degree for calculating a recommendation degree of each content for the user using the similarity calculated by the second similarity calculation means and the history information necessary for analysis from the log information for each session A calculation means;
    With
    The recommendation result output means outputs the recommendation degree for each content calculated by the first recommendation degree calculation means and the content for each content calculated by the second recommendation degree calculation means according to the change result of the integration ratio calculation means. The recommendation apparatus according to claim 1, wherein a recommendation result is output from the recommendation level.
  4. 　前記長期的嗜好分析処理手段が、
    　すべてのログ情報の中から解析に必要な履歴情報を収集する第１の履歴情報収集手段と、
    　該収集した履歴情報を解析して、ユーザ間の類似性を定義する第１の類似性算出手段と、
    　該算出された類似性と該セッションごとのログ情報の中から解析に必要な履歴情報とを用いて、ユーザに対する各コンテンツの推薦度を算出する第１の推薦度算出手段と、
    　該定義した類似性および算出した推薦度を記憶する記憶手段と、
    　を備え、
    　前記短期的興味分析処理手段が、
    　セッションごとのログ情報の中から解析に必要な履歴情報をリアルタイムに収集する第２の履歴情報収集手段と、
    　前記定義した類似性および算出した推薦度と前記セッションごとのログ情報の中から解析に必要な履歴情報とを用いて、ユーザに対する各コンテンツの推薦度を算出する第２の推薦度算出手段と、
    　前記レコメンド結果出力手段が、前記統合割合算出手段の変更結果に応じて、前記第１の推薦度算出手段により算出したコンテンツごとの推薦度と前記第２の推薦度算出手段により算出したコンテンツごとの推薦度とからレコメンド結果を出力することを特徴とする請求項１に記載のレコメンド装置。     The long-term preference analysis processing means is
    A first history information collecting means for collecting history information necessary for analysis from all log information;
    Analyzing the collected history information and defining first similarity between users; and
    First recommendation degree calculating means for calculating a recommendation degree of each content for the user using the calculated similarity and history information necessary for analysis from the log information for each session;
    Storage means for storing the defined similarity and the calculated recommendation degree;
    With
    The short-term interest analysis processing means includes:
    A second history information collecting means for collecting history information necessary for analysis from log information for each session in real time;
    Second recommendation degree calculation means for calculating a recommendation degree of each content for the user using the defined similarity and the calculated recommendation degree and history information necessary for analysis from the log information for each session;
    The recommendation result output means outputs the recommendation degree for each content calculated by the first recommendation degree calculation means and the content for each content calculated by the second recommendation degree calculation means according to the change result of the integration ratio calculation means. The recommendation apparatus according to claim 1, wherein a recommendation result is output from the recommendation level.
  5. 　前記セッションごとのログ情報を格納するデータベースと、
    　前記レコメンド結果出力手段から出力されるレコメンドについて、ユーザの閲覧状況を取得する閲覧状況取得手段と、
    　該取得したユーザの閲覧状況を前記データベースと前記統合割合算出手段にフィードバックする閲覧状況フィードバック手段と、
    　を備え、
    　前記統合割合算出手段が閲覧状況フィードバック情報に基づいて、前記レコメンド結果出力手段のレコメンド結果をユーザの興味の遷移に基づいて変更することを特徴とする請求項１から請求項４に記載のレコメンド装置。     A database for storing log information for each session;
    About the recommendation output from the recommendation result output means, browsing status acquisition means for acquiring the browsing status of the user,
    Browsing status feedback means for feeding back the acquired browsing status of the user to the database and the integrated ratio calculation means;
    With
    5. The recommendation device according to claim 1, wherein the integrated ratio calculation unit changes a recommendation result of the recommendation result output unit based on a transition of a user's interest based on browsing status feedback information. .
  6. 　購買履歴や閲覧履歴等のユーザの履歴情報を元にユーザ間の類似性を定義し、該定義した類似性に基づいてユーザが興味を持つと予測されるコンテンツをレコメンド結果として出力する協調フィルタリング方法を用いるレコメンドシステムであって、
    　すべての前記履歴情報を格納する第１のデータベースと、
    　セッションごとの前記履歴情報を格納する第２のデータベースと、
    　前記第１のデータベースに格納された履歴情報を解析してレコメンドに必要な情報を生成する長期的嗜好分析処理装置と、
    　前記第２のデータベースに格納されたセッションごとの履歴情報を解析してレコメンドに必要な情報を生成する短期的興味分析処理装置と、
    　前記長期的嗜好分析処理装置の出力または／および前記短期的興味分析処理装置の出力を統合してレコメンド結果を出力するレコメンド結果出力装置と、
    　前記レコメンド結果出力装置のレコメンド結果をユーザの興味の遷移に基づいて変更する統合割合算出装置と、
    　を備えたことを特徴とするレコメンドシステム。     A collaborative filtering method that defines similarity between users based on user history information such as purchase history and browsing history, and outputs content predicted to be of interest to the user based on the defined similarity as a recommendation result A recommendation system using
    A first database storing all the history information;
    A second database for storing the history information for each session;
    A long-term preference analysis processing device that analyzes history information stored in the first database and generates information necessary for recommendation;
    A short-term interest analysis processing device that analyzes history information for each session stored in the second database and generates information necessary for recommendation;
    A recommendation result output device for outputting a recommendation result by integrating the output of the long-term preference analysis processing device and / or the output of the short-term interest analysis processing device;
    An integrated ratio calculation device that changes a recommendation result of the recommendation result output device based on a transition of interest of the user;
    The recommendation system characterized by having.
  7. 　前記長期的嗜好分析処理装置が、
    　すべてのログ情報の中から解析に必要な履歴情報を収集する第１の履歴情報収集手段と、
    　該収集した履歴情報を解析して、ユーザ間の類似性を定義する第１の類似性算出手段と、
    　該算出された類似性と該すべてのログ情報の中から収集された解析に必要な履歴情報とを用いて、ユーザに対する各コンテンツの推薦度を算出する第１の推薦度算出手段と、
    　該算出した推薦度を記憶する記憶手段と、
    　を備え、
    　前記短期的興味分析処理装置が、
    　セッションごとのログ情報の中から解析に必要な履歴情報をリアルタイムに収集する第２の履歴情報収集手段と、
    　該収集した履歴情報を解析して、ユーザ間の類似性を定義する第２の類似性算出手段と、
    　該算出された類似性と該セッションごとのログ情報の中から解析に必要な履歴情報とを用いて、ユーザに対する各コンテンツの推薦度を算出する第２の推薦度算出手段と、
    　を備え、
    　前記レコメンド結果出力装置が、前記統合割合算出装置の変更結果に応じて、前記記憶されたコンテンツごとの推薦度と前記第２の推薦度算出手段により算出したコンテンツごとの推薦度とからレコメンド結果を出力することを特徴とする請求項６に記載のレコメンドシステム。     The long-term preference analysis processing device,
    A first history information collecting means for collecting history information necessary for analysis from all log information;
    Analyzing the collected history information and defining first similarity between users; and
    First recommendation degree calculation means for calculating a recommendation degree of each content for the user using the calculated similarity and history information necessary for analysis collected from all the log information;
    Storage means for storing the calculated recommendation degree;
    With
    The short-term interest analysis processing device comprises:
    A second history information collecting means for collecting history information necessary for analysis from log information for each session in real time;
    A second similarity calculation means for analyzing the collected history information and defining the similarity between users;
    Second recommendation degree calculating means for calculating a recommendation degree of each content for the user using the calculated similarity and history information necessary for analysis from the log information for each session;
    With
    The recommendation result output device obtains a recommendation result from the stored recommendation level for each content and the recommendation level for each content calculated by the second recommendation level calculation means according to the change result of the integrated ratio calculation device. The recommendation system according to claim 6, wherein the recommendation system is output.
  8. [規則26に基づく補充 10.01.2013]　
    　前記長期的嗜好分析処理装置が、
    　すべてのログ情報の中から解析に必要な履歴情報を収集する第１の履歴情報収集手段と、
    　該収集した履歴情報を解析して、ユーザ間の類似性を定義する第１の類似性算出手段と、
    　該算出した類似性を記憶する記憶手段と、
    　を備え、
    　前記短期的興味分析処理装置が、
    　セッションごとのログ情報の中から解析に必要な履歴情報をリアルタイムに収集する第２の履歴情報収集手段と、
    　該収集した履歴情報を解析して、ユーザ間の類似性を定義する第２の類似性算出手段と、
    　該セッションごとのログ情報の中から解析に必要な履歴情報と前記記憶された類似性とを用いて、ユーザに対する各コンテンツの推薦度を算出する第１の推薦度算出手段と、
    　前記第２の類似性算出手段により算出された類似性と前記セッションごとのログ情報の中から解析に必要な履歴情報とを用いて、ユーザに対する各コンテンツの推薦度を算出する第２の推薦度算出手段と、
    　を備え、
    　前記レコメンド結果出力装置が、前記統合割合算出装置の変更結果に応じて、前記第１の推薦度算出手段により算出したコンテンツごとの推薦度と前記第２の推薦度算出手段により算出したコンテンツごとの推薦度とからレコメンド結果を出力することを特徴とする請求項６に記載のレコメンドシステム。     [Supplement under rule 26 10.01.2013]
    The long-term preference analysis processing device,
    A first history information collecting means for collecting history information necessary for analysis from all log information;
    Analyzing the collected history information and defining first similarity between users; and
    Storage means for storing the calculated similarity;
    With
    The short-term interest analysis processing device comprises:
    A second history information collecting means for collecting history information necessary for analysis from log information for each session in real time;
    A second similarity calculation means for analyzing the collected history information and defining the similarity between users;
    First recommendation degree calculating means for calculating a recommendation degree of each content for a user using history information necessary for analysis from the log information for each session and the stored similarity;
    A second recommendation degree for calculating a recommendation degree of each content for the user using the similarity calculated by the second similarity calculation means and the history information necessary for analysis from the log information for each session A calculation means;
    With
    The recommendation result output device determines the recommendation level for each content calculated by the first recommendation level calculation unit and the content for each content calculated by the second recommendation level calculation unit according to the change result of the integration ratio calculation device. The recommendation system according to claim 6, wherein a recommendation result is output from the recommendation level.
  9. 　前記長期的嗜好分析処理装置が、
    　すべてのログ情報の中から解析に必要な履歴情報を収集する第１の履歴情報収集手段と、
    　該収集した履歴情報を解析して、ユーザ間の類似性を定義する第１の類似性算出手段と、
    　該算出された類似性該セッションごとのログ情報の中から解析に必要な履歴情報とを用いて、ユーザに対する各コンテンツの推薦度を算出する第１の推薦度算出手段と、
    　該定義した類似性および算出した推薦度を記憶する記憶手段と、
    　を備え、
    　前記短期的興味分析処理装置が、
    　セッションごとのログ情報の中から解析に必要な履歴情報をリアルタイムに収集する第２の履歴情報収集手段と、
    　前記定義した類似性および算出した推薦度と前記セッションごとのログ情報の中から解析に必要な履歴情報とを用いて、ユーザに対する各コンテンツの推薦度を算出する第２の推薦度算出手段と、
    　を備え、
    　前記レコメンド結果出力装置が、前記統合割合算出装置の変更結果に応じて、前記第１の推薦度算出手段により算出したコンテンツごとの推薦度と前記第２の推薦度算出手段により算出したコンテンツごとの推薦度とからレコメンド結果を出力することを特徴とする請求項６に記載のレコメンドシステム。     The long-term preference analysis processing device,
    A first history information collecting means for collecting history information necessary for analysis from all log information;
    Analyzing the collected history information and defining first similarity between users; and
    A first recommendation degree calculating means for calculating a recommendation degree of each content for the user using history information necessary for analysis from the calculated log information for each session;
    Storage means for storing the defined similarity and the calculated recommendation degree;
    With
    The short-term interest analysis processing device comprises:
    A second history information collecting means for collecting history information necessary for analysis from log information for each session in real time;
    Second recommendation degree calculation means for calculating a recommendation degree of each content for the user using the defined similarity and the calculated recommendation degree and history information necessary for analysis from the log information for each session;
    With
    The recommendation result output device determines the recommendation level for each content calculated by the first recommendation level calculation unit and the content for each content calculated by the second recommendation level calculation unit according to the change result of the integration ratio calculation device. The recommendation system according to claim 6, wherein a recommendation result is output from the recommendation level.
  10. [規則26に基づく補充 10.01.2013]　
    　出力するレコメンドについて、ユーザの閲覧状況を取得する閲覧状況取得装置と、
    　該取得したユーザの閲覧状況を前記第２のデータベースと前記統合割合算出手段にフィードバックする閲覧状況フィードバック装置と、
    　を備え、
    　前記統合割合算出装置が閲覧状況フィードバック情報に基づいて、前記レコメンド結果出力装置のレコメンド結果をユーザの興味の遷移に基づいて変更することを特徴とする請求項６から請求項９に記載のレコメンドシステム。     [Supplement under rule 26 10.01.2013]
    About the recommendation to output, the browsing status acquisition device which acquires the browsing status of the user,
    A browsing status feedback device that feeds back the acquired browsing status of the user to the second database and the integrated ratio calculation means;
    With
    10. The recommendation system according to claim 6, wherein the integrated ratio calculation device changes a recommendation result of the recommendation result output device based on a transition of a user's interest based on browsing status feedback information. .
  11. 　購買履歴や閲覧履歴等のユーザの履歴情報を元にユーザ間の類似性を定義し、該定義した類似性に基づいてユーザが興味を持つと予測されるコンテンツをレコメンド結果として出力する協調フィルタリング方法を用いるレコメンド方法であって、
    　すべての前記履歴情報を解析してレコメンドに必要な情報を生成する長期的嗜好分析処理ステップと、
    　セッションごとの履歴情報を解析してレコメンドに必要な情報を生成する短期的興味分析処理ステップと、
    　前記長期的嗜好分析処理ステップの出力または／および前記短期的興味分析処理ステップの出力を統合してレコメンド結果を出力するレコメンド結果出力ステップと、
    　前記レコメンド結果出力ステップのレコメンド結果をユーザの興味の遷移に基づいて変更する統合割合算出ステップと、
    　を備えたことを特徴とするレコメンド方法。     A collaborative filtering method that defines similarity between users based on user history information such as purchase history and browsing history, and outputs content predicted to be of interest to the user based on the defined similarity as a recommendation result A recommendation method using
    A long-term preference analysis processing step of analyzing all the history information and generating information necessary for recommendation;
    A short-term interest analysis processing step that analyzes the historical information for each session and generates the information required for the recommendation,
    A recommendation result output step of outputting a recommendation result by integrating the output of the long-term preference analysis processing step and / or the output of the short-term interest analysis processing step;
    An integration ratio calculating step for changing the recommendation result of the recommendation result output step based on a transition of the user's interest;
    A recommendation method characterized by comprising:
  12. 　購買履歴や閲覧履歴等のユーザの履歴情報を元にユーザ間の類似性を定義し、該定義した類似性に基づいてユーザが興味を持つと予測されるコンテンツをレコメンド結果として出力する協調フィルタリング方法を用いるレコメンド方法をコンピュータに実行させるためのプログラムであって、
    　すべての前記履歴情報を解析してレコメンドに必要な情報を生成する長期的嗜好分析処理ステップと、
    　セッションごとの履歴情報を解析してレコメンドに必要な情報を生成する短期的興味分析処理ステップと、
    　前記長期的嗜好分析処理ステップの出力または／および前記短期的興味分析処理ステップの出力を統合してレコメンド結果を出力するレコメンド結果出力ステップと、
    　前記レコメンド結果出力ステップのレコメンド結果をユーザの興味の遷移に基づいて変更する統合割合算出ステップと、
    　をコンピュータに実行させるためのプログラム。     A collaborative filtering method that defines similarity between users based on user history information such as purchase history and browsing history, and outputs content predicted to be of interest to the user based on the defined similarity as a recommendation result A program for causing a computer to execute a recommendation method using
    A long-term preference analysis processing step for analyzing all the history information and generating information necessary for recommendation;
    A short-term interest analysis processing step that analyzes the historical information for each session and generates the information required for the recommendation,
    A recommendation result output step of outputting a recommendation result by integrating the output of the long-term preference analysis processing step and / or the output of the short-term interest analysis processing step;
    An integration ratio calculating step for changing the recommendation result of the recommendation result output step based on a transition of the user's interest;
    A program that causes a computer to execute.
PCT/JP2012/080921 2011-12-02 2012-11-29 Recommendation device, recommendation system, recommendation method and program WO2013081051A1 (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201280058706.2A CN104246751B (en) 2011-12-02 2012-11-29 Recommendation apparatus, commending system and recommendation method

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
JP2011264625A JP5798022B2 (en) 2011-12-02 2011-12-02 RECOMMENDATION DEVICE, RECOMMENDATION SYSTEM, RECOMMENDATION METHOD, AND PROGRAM
JP2011-264625 2011-12-02

Publications (1)

Publication Number Publication Date
WO2013081051A1 true WO2013081051A1 (en) 2013-06-06

Family

ID=48535500

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/JP2012/080921 WO2013081051A1 (en) 2011-12-02 2012-11-29 Recommendation device, recommendation system, recommendation method and program

Country Status (3)

Country Link
JP (1) JP5798022B2 (en)
CN (1) CN104246751B (en)
WO (1) WO2013081051A1 (en)

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106997358A (en) * 2016-01-22 2017-08-01 中移（杭州）信息技术有限公司 Information recommendation method and device
CN110929164A (en) * 2019-12-09 2020-03-27 北京交通大学 Interest point recommendation method based on user dynamic preference and attention mechanism
CN112163147A (en) * 2020-06-09 2021-01-01 中森云链(成都)科技有限责任公司 Recommendation method for website session scene
CN112905887A (en) * 2021-02-22 2021-06-04 中国计量大学 Conversation recommendation method based on multi-interest short-term priority model
CN112948683A (en) * 2021-03-16 2021-06-11 山西大学 Socialized recommendation method with dynamic fusion of social information

Families Citing this family (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2015060547A (en) * 2013-09-20 2015-03-30 シャープ株式会社 Information processing device, information processing system, information processing method, information processing program, and terminal device
CN106033413A (en) * 2015-03-09 2016-10-19 阿里巴巴集团控股有限公司 An information base generation method and device and an information search method
CN106503014B (en) 2015-09-08 2020-08-07 腾讯科技（深圳）有限公司 Real-time information recommendation method, device and system
CN107093095B (en) * 2016-12-02 2018-05-25 口碑(上海)信息技术有限公司 Method and device is recommended in associated services processing method and processing device, shop
CN109034972A (en) * 2018-07-24 2018-12-18 合肥爱玩动漫有限公司 Commodity method for pushing based on player preferences in a kind of game
CN111125537B (en) * 2019-12-31 2020-12-22 中国计量大学 Session recommendation method based on graph representation
CN113535311A (en) * 2021-07-29 2021-10-22 展讯半导体(成都)有限公司 Page display method and device and electronic equipment

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2004326227A (en) * 2003-04-22 2004-11-18 Matsushita Electric Ind Co Ltd Information providing method, information providing system, its program, and program storage medium
EP1873657A1 (en) * 2006-06-29 2008-01-02 France Télécom User-profile based web page recommendation system and method

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP5032183B2 (en) * 2007-04-12 2012-09-26 株式会社東芝 Information recommendation system and information recommendation method
CN101436186B (en) * 2007-11-12 2012-09-05 北京搜狗科技发展有限公司 Method and system for providing related searches
JP2011145742A (en) * 2010-01-12 2011-07-28 Sony Corp Apparatus and method for processing information, and program
CN101968802A (en) * 2010-09-30 2011-02-09 百度在线网络技术（北京）有限公司 Method and equipment for recommending content of Internet based on user browse behavior

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2004326227A (en) * 2003-04-22 2004-11-18 Matsushita Electric Ind Co Ltd Information providing method, information providing system, its program, and program storage medium
EP1873657A1 (en) * 2006-06-29 2008-01-02 France Télécom User-profile based web page recommendation system and method

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
KAZUNARI SUGIYAMA: "Adaptive Web search based on a word-based collaborative filtering that overcomes data sparsity", DAI 67 KAI JINKO CHINO KIHON MONDAI KENKYUKAI SHIRYO (SIG-FPAI- A702), 25 October 2007 (2007-10-25), pages 7 - 12 *
RAFTER, R. ET AL.: "Conversational Collaborative Recommendation - An Experimental Analysis", THE ARTIFICIAL INTELLIGENCE REVIEW, vol. 24, no. 3-4, November 2005 (2005-11-01), pages 301 - 318, XP019390040 *
TOSHIHIRO KAMISHIMA: "Algorithms for Recommender Systems (3)", JOURNAL OF JAPANESE SOCIETY FOR ARTIFICIAL INTELLIGENCE, vol. 23, no. 2, 1 March 2008 (2008-03-01), pages 248 - 263, XP008171428 *

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106997358A (en) * 2016-01-22 2017-08-01 中移（杭州）信息技术有限公司 Information recommendation method and device
CN110929164A (en) * 2019-12-09 2020-03-27 北京交通大学 Interest point recommendation method based on user dynamic preference and attention mechanism
CN110929164B (en) * 2019-12-09 2023-04-21 北京交通大学 Point-of-interest recommendation method based on user dynamic preference and attention mechanism
CN112163147A (en) * 2020-06-09 2021-01-01 中森云链(成都)科技有限责任公司 Recommendation method for website session scene
CN112905887A (en) * 2021-02-22 2021-06-04 中国计量大学 Conversation recommendation method based on multi-interest short-term priority model
CN112905887B (en) * 2021-02-22 2021-12-14 中国计量大学 Conversation recommendation method based on multi-interest short-term priority model
CN112948683A (en) * 2021-03-16 2021-06-11 山西大学 Socialized recommendation method with dynamic fusion of social information
CN112948683B (en) * 2021-03-16 2022-11-11 山西大学 Social recommendation method with dynamic fusion of social information

Also Published As

Publication number Publication date
JP5798022B2 (en) 2015-10-21
CN104246751A (en) 2014-12-24
CN104246751B (en) 2017-05-31
JP2013117830A (en) 2013-06-13

Similar Documents

Publication Publication Date Title
JP5798022B2 (en) RECOMMENDATION DEVICE, RECOMMENDATION SYSTEM, RECOMMENDATION METHOD, AND PROGRAM
US10650316B2 (en) Issue-manage-style internet public opinion information evaluation management system and method thereof
CA2844130C (en) Filtering social search results
CN107784066A (en) Information recommendation method, device, server and storage medium
US11057328B2 (en) Real-time recommendation of message recipients based on recipient interest level in message
CN104462560A (en) Personalized recommendation system and method
TW201214167A (en) Matching text sets
US20140229487A1 (en) System and method for user preference augmentation through social network inner-circle knowledge discovery
CN103136253A (en) Method and device of acquiring information
US20150213136A1 (en) Method and System for Providing a Personalized Search List
EP2943897A1 (en) Method and apparatus for enriching social media to improve personalized user experience
US20100318427A1 (en) Enhancing database management by search, personal search, advertising, and databases analysis efficiently using core-set implementations
US20140164274A1 (en) Apparatus and Method for Recommending Courses
CN111159341A (en) Information recommendation method and device based on user investment and financing preference
CN107103028A (en) A kind of information processing method and device
KR101621735B1 (en) Recommended search word providing method and system
KR101568800B1 (en) Real-time issue search word sorting method and system
JP5806081B2 (en) RECOMMENDATION DEVICE, RECOMMENDATION SYSTEM, RECOMMENDATION METHOD, AND PROGRAM
Sacco et al. In users we trust: towards social user interactions based Trust Assertions for the Social Semantic Web
Chiang et al. Exploring smartphone users’ social information behavior
US9384513B2 (en) Apparatus for collecting contents using social relation character and method thereof
Kiritoshi et al. A diversity-seeking mobile news app based on difference analysis of news articles
US20210382953A1 (en) Methods and systems for adjusting self-tuning personalization engines in near real-time
Zeller The Good, the Bad, and the How-to of Analysing Social Media Data
El Marrakchi et al. Intelligent reputation scoring in social networks: Use case of brands of smartphones

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 12852559

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 12852559

Country of ref document: EP

Kind code of ref document: A1