JP2017182746A - Information provision server device, program and information provision method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an information provision server, etc., with which it is possible to provide standard information which gives unexpectedness but is acceptable to a user.SOLUTION: The information provision server comprises: means for counting the number of items purchased, etc., for each category; means for calculating the degree of similarity between users; means for classifying other users for a request user into high, middle and low similarity classes; means for comparing the items purchased by the request user with the items purchased by users of high level similarity and extracting a category for which a prescribed relationship about the number of items purchased, etc., is established; means for finding the degree of standardness on the basis of the number of items purchased by users belonging to the middle and low level similarity classes from among the items included in the extracted category; means for extracting, on the basis of the degree of standardness, a prescribed item from among the items included in the extracted category; and means for providing information pertaining to the extracted item.SELECTED DRAWING: Figure 1

Description

  The present invention relates to an information providing server device or the like that provides information on a standard item to a user while having unexpectedness for the user.

  In recent years, with the development of information communication technology, a recommendation system for providing recommendation information to users for various items such as books and other products has become widespread. In the recommendation system currently in widespread use, many collaborative filtering systems and content-based filtering systems have been developed, and such recommendation systems have been disclosed so far (for example, Patent Document 1 to Patent Document 3). .

  In these systems, the information provided to the user may be of an unexpected nature that the user has not anticipated or, in some cases, a surprise. Since these properties are generally referred to as “unexpectedness”, in the following description, these properties will be described as “unexpectedness”.

  The system disclosed in Patent Document 1 is a system that recommends a document with a low category similarity and a high text similarity compared to the text selected by the user. Specifically, the document search system disclosed in Patent Document 1 first receives a selection of a search target document from a user terminal, and generates a document vector representing the characteristics of the search target document. Next, the similarity between the document group belonging to each category registered in the category information database and the search target document is calculated, and the similarity between each document registered in the document information DB 6 and the search target document is calculated. Then, the document search system searches for a document that belongs to a category that has a low similarity to the document selected by the user based on these, but whose contents are similar among the documents, and uses the searched document as a recommended destination document. Presenting to the terminal.

  The system disclosed in Patent Document 2 recommends information using a combination of both the same category and between different categories in the similarity calculation between a user profile and a book. Specifically, the system disclosed in Patent Document 2 first synthesizes a keyword vector of each book purchased by a customer in the past for each category to generate a customer profile. Further, a keyword vector is generated for each book based on the book database. Then, by calculating the similarity between the keyword vector of the customer profile and the keyword vector of each unpurchased book, both within the same category and between different categories, both beneficial and unexpected aspects Searching for books from.

  In the system disclosed in Patent Document 3, knowledge already acquired by users belonging to a plurality of communities is estimated, and information in unexpected fields is recommended. In other words, in the system disclosed in Patent Document 3, the similarity calculation unit handles information having affinity for a predetermined user by another user who belongs to the same user in one or more communities. The first index value representing the degree of presence is calculated from the characteristics of communities that belong in common and the characteristics of all communities to which a predetermined user belongs. In addition, the unexpectedness calculation means commonly uses a second index value that represents the degree to which other users who commonly belong to the predetermined user in one or more communities bring information in a field unexpected to the predetermined user. And the characteristics of the users belonging to the community to which the predetermined user does not belong and among the users belonging to the community to which the other user belongs. And the recommendation means selects other users who recommend information to a predetermined user based on the result of calculating the index values for each user.

Japanese Patent Laying-Open No. 2015-197722 JP 2001-265808 A JP 2010-113683 A

  However, when recommendation information is created using a system that uses a collaborative filtering method and a content-based filtering method, such as the systems disclosed in Patent Document 1 to Patent Document 3, information on different fields and different categories is touched. Users can feel surprised, but there is no particular context with the information that they usually use, such as books they normally read, and there is a risk that they may receive a sudden impression .

  The present invention has been made to solve the above-described problems, and an object of the present invention is to provide an information providing server and the like that can provide standard information that is easy to accept while the user feels unexpected. .

  (1) In order to solve the above-described problem, the information providing server device of the present invention is connected to a communication terminal device possessed by a user via a network, and recommends information regarding an item constituting one type of predetermined content to the user. An information providing server device for providing information to the communication terminal device as information, the counting means for counting the number of purchases or the number of browsing of the item purchased or viewed by each user for each category to which the item belongs, Based on the number of purchases or the number of times of browsing, similarity calculation means for calculating the degree of similarity of purchase or browsing with respect to the content among users, and the recommendation information based on the calculated similarity between the users Based on the threshold, other users with respect to the requesting user requesting a high degree of similarity class, medium similarity A user group classifying means for classifying into a user group belonging to three classes of similarity classes of a class and a low degree similarity class, and the item purchased or viewed by each user constituting the user group of the high degree similarity class, The item that the requesting user has already purchased or browsed is compared for each category, and a value obtained by quantifying the correlation between the number of purchases or the number of browsing for each category is predetermined between the number of purchases or the number of browsing for each category. An unexpected category extracting means for extracting a category having a relationship satisfying the above condition as an unexpected category, and the medium similarity class and the items included in the category extracted by the unexpected category extracting means, and Based on the number of purchases or the number of browsing of items purchased or viewed by users in the user group belonging to the low degree similarity class, The standard degree calculation means for obtaining a standard degree obtained by quantifying the index as the degree of recognition of each item, and the category extracted by the unexpected category extraction means based on the standard degree calculated by the standard degree calculation means An item extracting means for extracting an item having a predetermined standardity from the items included in the information, and an information providing stage for providing information relating to the item extracted by the item extracting means to the requesting user as the recommended information It has the composition to provide.

  With this configuration, among users who have similar preferences, purchases or browsing are rarely made even though they are in the same category, and purchases or browsing is increased with users who do not show much interest. It can be seen that there are users who show great interest. Then, by providing information on items included in a category in which there is a user who shows a great deal of interest to a user who does not show a lot of interest, the user can be surprised.

  On the other hand, an item having a history of purchase even by a user having a medium degree of similarity and a user having a low degree can be generally regarded as an item having a high standard degree. By providing the requesting user with information related to items with a high standardity, it is possible to provide information on items that are relatively easy to accept and not sudden to the user.

  Therefore, as described above, an unexpected category is extracted based on an item that has been purchased or viewed by a user with a high degree of similarity, and a standard item is based on an item that has been purchased or viewed by a user with a medium or low degree of similarity. By calculating the degree, it is possible to provide the requesting user with information about an item that is easy to accept while giving unexpectedness.

  (2) Further, in the information providing server device according to the present invention, the user group classifying unit is configured to assign a user of a user group belonging to the high degree similarity class, medium similarity class, and low degree similarity class to a predetermined degree related to similarity. Sorting means for sorting according to the above rule and sorting by the threshold value is provided for each class.

  With this configuration, when a predetermined number of users are extracted from user groups having different degrees of similarity, it is easy to extract users having similarity characteristics in each user group. For example, when the similarities of users in the high degree of similarity class are sorted in descending order and several names with the highest similarity are extracted, the extracted users have a characteristic that the similarity is high. .

  (3) Further, in the information providing server device of the present invention, the predetermined relationship in each category is such that the number of purchases or the number of browsing times that each user who constitutes a user group belonging to the high degree similarity class purchases an item The difference obtained by subtracting the number of purchases that the requesting user has already purchased or the number of times of browsing viewed from the requesting user is greater than or equal to a predetermined value.

  With this configuration, the larger the difference in the number of items purchased or browsed for items included in the same category, the more different the preference for the category among users. In this way, by using the difference in the number of purchased or browsed, it is possible to objectively determine the difference in preference for the predetermined category of the user.

  (4) Moreover, the information provision server apparatus of this invention is the user group which belongs to the said medium similarity class and the said low degree similarity class from the items contained in the category extracted by the said unexpected category extraction means. It has the structure further provided with the filtering means which excludes the item which has never been purchased or browsed by the user.

  With this configuration, it is possible to avoid providing information relating to an unexpected item to the requesting user. That is, an item that has never been purchased or viewed by a user of a user group belonging to the medium similarity class and the low similarity class is a special (specialized) item that has low public awareness. it is conceivable that. This configuration can eliminate special (specialized) items with low public awareness.

  (5) Moreover, in the information providing server device of the present invention, the standard degree calculating means calculates the standard degree by calculating the number of items purchased or the number of browsed for each item. Have.

  With this configuration, it is possible to distinguish items with a high standard degree and items with a low standard degree, or to rank items with a standard degree. As a result, it is possible to provide information on items having a high standardity to the requesting user.

  (6) Further, in the information providing server device of the present invention, the item extraction means is an item that the requesting user has not purchased or browsed yet, and belongs to a user group of the high degree of similarity class Has a configuration for extracting items that have already been purchased or already viewed.

  With this configuration, the requesting user can provide information about items in a category that has not shown much interest until then, and the requesting user can be surprised.

  (7) In order to solve the above-described problem, a program according to the present invention is connected to a communication terminal device possessed by a user via a network, and uses information related to an item that constitutes one type of predetermined content for the user as recommendation information. A program used in a computer constituting an information providing server device for providing to the communication terminal device, wherein the item indicates the number of purchases or the number of browsing times of the item purchased or viewed by each user. Counting means for counting for each category to which the user belongs, similarity calculating means for calculating a degree of purchase or viewing preference for the content among users based on the number of purchases or the number of browsing, and the similarity between the calculated users Based on the degree to the requesting user requesting the recommendation information User group classification means for classifying other users into user groups belonging to three categories of similarity classes of a high degree of similarity class, a medium degree of similarity class, and a low degree of similarity class based on a threshold, the high degree of similarity The item purchased or browsed by each user constituting the user group of the degree class and the item already purchased or browsed by the requesting user are compared for each category, and between the number of purchases or the number of browsing for each category Extracted by the unexpected category extracting means, the unexpected category extracting means for extracting, as the unexpected category, a category in which the relationship in which the value obtained by quantifying the correlation between the number of purchases or the number of browsing for each category satisfies a predetermined condition is established Among the items included in the selected category, by users of the user group belonging to the medium similarity class and the low similarity class Based on the number of purchases or the number of browsing of items that have been entered or browsed, the standard degree calculation means for obtaining a standard degree obtained by quantifying an index as the degree of recognition of each item, the standard degree calculated by the standard degree calculation means Based on the item extraction means for extracting an item having a predetermined standardity from the items included in the category extracted by the unexpected category extraction means, and information on the item extracted by the item extraction means It has a configuration that functions as an information provision stage provided to the requesting user as recommendation information.

  With this configuration, an unexpected category is extracted based on an item that has been purchased or viewed by a user with a high degree of similarity, and a standard degree is calculated based on an item purchased or viewed by a user with a medium or low degree of similarity. Thus, it is possible to provide the requesting user with information about items that are easy to accept while giving unexpectedness.

  (8) In order to solve the above-described problem, an information providing method according to the present invention is connected to a communication terminal device possessed by a user via a network, and is information on an item that constitutes one type of predetermined content for the user. An information providing method for providing to the communication terminal device, the counting step of counting the number of purchases or the number of browsing of the item purchased or browsed by each user for each category to which the item belongs, and the number of purchases Alternatively, the recommendation information is requested based on the similarity calculation step of calculating the similarity regarding the purchase or browsing preference for the content among the users based on the number of times of browsing and the calculated similarity between the users. Based on the threshold, other users with respect to the requesting user are classified as high similarity class, medium similarity class and low A user group dividing step of dividing into three user groups belonging to the similarity class of the similarity class, the item purchased or viewed by each user constituting the user group of the high degree similarity class, and the requesting user The item that has already been purchased or viewed is compared for each category, and the value obtained by quantifying the correlation between the number of purchases or the number of browsing for each category has a predetermined condition between the number of purchases or the number of browsing for each category. An unexpected category extraction step for extracting a category in which the relationship is established as an unexpected category, and the intermediate similarity class and the lower similarity from the items included in the category extracted by the unexpected category extraction unit Based on the number of purchases or browsing times of items purchased or viewed by users in the user group belonging to the degree class, A standard frequency calculation step for obtaining a standard frequency obtained by quantifying an index as a degree of knowledge, and an item included in the category extracted by the unexpected category extraction unit based on the standard frequency calculated by the standard frequency calculation unit An item extracting step for extracting an item having a predetermined standardity from the above, and an information providing step for providing information relating to the item extracted by the item extracting means to the requesting user as the recommended information. doing.

  With this configuration, unexpected categories are extracted based on items that have been purchased or viewed by users with a high degree of similarity, and standardity is calculated based on items that have been purchased or viewed by users with a medium or low degree of similarity. By doing so, it is possible to provide information about an item that can be accepted easily while giving the requesting user an unexpectedness.

  The information providing server device and the like according to the present invention can provide standard information that is easy to accept while the user feels unexpected.

It is a system configuration figure showing the composition of the information service system in one embodiment of the present invention. It is a block diagram which shows the structure of the information provision server apparatus in one Embodiment. It is a figure which shows an example of the data registered into the user information database in the information provision server apparatus in one Embodiment. It is a figure which shows an example of the data registered into the content information database in the information provision server apparatus in one Embodiment. It is a figure which shows an example of the data registered into the category information database in the information provision server apparatus in one Embodiment. In the information provision system in one Embodiment, it is a schematic diagram which shows the outline | summary of the whole flow of the system which provides recommendation information to a requesting user. It is explanatory drawing for demonstrating the principle which extracts the user with similar preference in the information provision server apparatus in one Embodiment. In the information provision server apparatus in one Embodiment, it is explanatory drawing for demonstrating the extraction function of the item contained in a predetermined category and giving a user an unexpectedness. It is explanatory drawing for demonstrating the calculation process of a standardity in the information provision server apparatus in one Embodiment. It is a flowchart which shows the process performed by the information provision server apparatus in one Embodiment.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings. In the following embodiment, the communication terminal device 10 owned by each user and the communication terminal device 10 are connected to the network 20 directly or via the base station BS to manage each communication terminal device 10 and to recommend information. This is an embodiment when applied to the information providing server device, the program, and the information providing system 1 according to the present invention, based on a system including the information providing server device 30 that provides the communication terminal device 10 with the information providing server device 30.

[1] Overview of Information Providing System First, the information providing system 1 according to the present embodiment will be described with reference to FIG. FIG. 1 is a system configuration diagram showing the configuration of the information providing system 1 of the present embodiment. In order to prevent the figure from becoming complicated, only a part of the user is shown in FIG. That is, in the actual information providing system 1, there are a larger number of communication terminal devices 10 than those displayed in FIG.

  The information providing system 1 according to the present embodiment is connected to a communication terminal device 10 possessed by a user via a network 20 and provides an information providing server device for providing information relating to an item suitable for each user to the user as recommended information. 30 is a system that provides recommendation information suitable for the user. In particular, the information providing system 1 according to the present embodiment relates to a standard item that is not only surprising but can be accepted without a sense of resistance for a user who requests recommendation information (hereinafter referred to as “requesting user”). It is characterized by providing information.

  In the present embodiment, the content to be handled is described using an electronic book as an example, but the content that can be handled by the information providing system 1 of the present embodiment is not limited to the electronic book. Examples of contents that can be handled include electronic books and contents that can reflect hobbies such as music and recommended gourmets.

  Specifically, the information providing system 1 according to the present embodiment is connected to the communication terminal device 10, the network 20, and the communication terminal device 10 of each user directly or via the base station BS. The information providing server device 30 that manages each communication terminal device 10 is provided.

  Moreover, the information provision system 1 in this embodiment employs a collaborative filtering method. That is, preference information for many users' contents is accumulated, and one user purchases or browses an item such as a book that constitutes a certain content, and another user purchases or browses the data. It is a system that recommends personalized information to one user by analyzing the correlation between the users based on the patterns purchased and viewed by the user using both of the data.

  As described above, the information providing system 1 according to the present embodiment is configured based on the collaborative filtering method and does not employ the content-based filtering method. Therefore, the system can have diversity and versatility.

  As the communication terminal device 10 used in the present embodiment, for example, a personal computer (PC), a smartphone, a tablet information terminal device, a mobile phone, a game machine, a PDA, or the like is used.

  The communication terminal device 10 uses resource data for the WWW system described in a markup language such as XML (extensible Markup Language), and a unique URL (Uniform Resource Locator) indicating a network address of the resource data, A browsing function for performing data communication with the information providing server device 30 is provided. Specifically, the communication terminal device 10 displays various contents (including work content) constituted by the resource data received from the information providing server device 30 using a browsing function so that the user can view them, By providing communication connection with the information providing server device 30 via the base station BS and the network 20, it is possible to receive various SNS services from the information providing server device 30.

  The information providing server device 30 is an information providing server device used for providing each service of SNS. The information providing server device 30 has a database (hereinafter referred to as “DB”) and cooperates with the communication terminal device 10 based on a user instruction input via the communication terminal device 10 possessed by each user. However, a process of registering various information in the DB is executed.

The information providing server device 30 has a configuration for executing the following functions based on the above configuration. That is, the information providing server device 30
(1) A function of counting the number of purchases or the number of browsing of items purchased or viewed by each user for each category to which the item belongs,
(2) A function of calculating a similarity degree regarding purchase or browsing preferences for content between users based on the number of purchases or the number of browsing,
(3) Based on the calculated similarity between the users, other users with respect to the user who requests the recommendation information are assigned the high similarity class, the intermediate similarity class, and the low similarity class based on the threshold. A function to divide into three groups of users belonging to the similarity class,
(4) The items purchased or browsed by each user constituting the user group of the high degree of similarity class and the items already purchased or browsed by the requesting user are compared for each category, and the number of purchases or the number of times of browsing for each category A function for extracting as a surprising category a category in which a relationship in which a value obtained by quantifying the correlation between the number of purchases or the number of browsing for each category satisfies a predetermined condition
(5) From the items included in the extracted category, each of the items purchased or viewed by the users of the user group belonging to the medium similarity class and the low similarity class, A means of finding a standardized measure that quantifies an indicator of the degree of item recognition,
(6) A function of extracting an item having a predetermined standardity from items included in the category extracted by the means for extracting the unexpected category based on the standardity calculated by the standardity calculation unit, as well as,
(7) A configuration for realizing a function of providing information relating to the item extracted by the means for extracting an item to the requesting user as recommendation information is provided.

  With the configuration for realizing the above-described function, the information providing system 1 according to the present embodiment has an effect specific to the system according to the present embodiment that allows the user to provide standard information that is easy to accept while feeling unexpected. Play.

[2] Information Providing Server Device Next, the configuration of the information providing server device 30 of the present embodiment will be described with reference to FIG. FIG. 2 is a block diagram showing the configuration of the information providing server device 30 of this embodiment.

  As shown in FIG. 2, the information providing server device 30 of the present embodiment functions as a server management control unit 310 that controls the entire device, a communication control unit 320 that communicates with each communication terminal device 10, and various memories. A ROM / RAM 330, a recording unit 340 that stores various types of information, and a data processing unit 350 that executes various types of processing, and the above-described units are connected to each other via a bus B.

  The server management control unit 310 is mainly configured by a central processing unit (CPU), and performs integrated control of each unit of the information providing server device 30 by executing a program.

  The communication control unit 320 is a predetermined network interface, and exchanges data with each communication terminal device 10 via the network 20. In addition, the communication control unit 320 cooperates with an information providing unit 358 described later to transmit recommendation information to the communication terminal device 10.

  Various programs necessary for driving the information providing server device 30 are recorded in the ROM / RAM 330. The ROM / RAM 330 is used as a work area when the server management control unit 310 and the data processing unit 350 execute various processes.

[2.1] Recording Unit The recording unit 340 is configured by, for example, an HDD (Hard Disk Drive) or an SSD (Solid State Drive).

  The recording unit 340 has at least the following DBs 341 to 343 in the recording area. That is, the recording unit 340 includes a user information DB 341 in which basic information related to a user who uses the information providing system 1 is registered in association with a user ID, a content information DB 342 in which information related to content purchased or viewed by the user is registered, And a category information DB 343 in which information on a category to which a book as an item constituting the content belongs is registered.

[2.2] Data Processing Unit The data processing unit 350 is configured by an independent central processing unit (CPU) or by using the central processing unit (CPU) of the server management control unit 310. Then, the data processing unit 350 manages processes executed in each communication terminal device 10 by executing a program under the control of the server management control unit 310.

  Specifically, the data processing unit 350 executes a program to manage a DB, a DB management unit 351, a count unit 352 that counts the number of books purchased or viewed for each category, and between users. , A similarity calculation unit 353 that calculates the degree of similarity to content, a user group classification unit 354 that classifies the degree of similarity between users into high, medium, and low user groups, and purchase for each category between users Or an unexpected category extraction unit 355 for extracting a category that is surprising to the user based on the relationship established between the number of times of browsing, the standard degree of the book based on the number of purchases or the number of times of browsing of the book constituting the content The standard degree calculation unit 356 for obtaining the item, the item extraction unit 357 for extracting a book having a predetermined standard degree, and the extracted book The information providing unit 358 to be provided to the requesting user, to realize.

(DB management department)
The DB management unit 351 constitutes a first database management unit, a second database management unit, and the like according to the present invention. The DB management unit 351 executes processing for registering various information in each DB constituting the recording unit 340, processing for searching for information registered in each DB, and processing for reading out information registered in each DB. .

  The DB management unit 351 executes, for the user information DB 341, processing for registering new information when a user is newly added, processing for searching for a user who has logged into the system, processing for reading the search result, and the like. To do.

  In addition, the DB management unit 351 registers, in the content information DB 342, a process of registering new information when a new book is added, a process of searching for information on the registered book, and a process of reading the searched information Etc.

  Further, the DB management unit 351 executes processing such as category registration, category search, and category reading on the category information DB 343.

(Counting part)
The counting unit 352 constitutes the counting means of the present invention. The counting unit 352 executes a process of counting the number of purchases or the number of browsing of books purchased or viewed by each user for each category to which the book belongs. When the item is a book, the category referred to here can include, for example, the fields of SF, mystery, IT, and business.

(Similarity calculator)
The similarity calculation unit 353 constitutes the similarity calculation means of the present invention. Based on the number of purchases or the number of browsing times, the similarity calculation unit 353 calculates the degree of similarity related to the purchase or browsing preferences for the content among the users. The method for calculating the similarity is not particularly limited. Similarity can be calculated by creating a vector for preference for the category of books purchased or viewed by each user and using this vector to calculate cosine similarity, or using a calculation formula using Euclidean distance. To calculate.

(User group division)
The user group sorting unit 354 constitutes the user group sorting means of the present invention. The user group classifying unit 354 assigns other users to the requesting user requesting recommendation information to user groups belonging to three categories of similarity classes of a high degree of similarity class, a medium degree of similarity class, and a low degree of similarity class. Sort. In addition, each similarity class may be classified by providing a numerical value that serves as a boundary in order to classify the classes, or may be classified based on the relative relationship of the calculated similarities. . However, based on the calculated similarity between users, the user group classifying unit 354 determines other users with respect to the requesting user requesting recommendation information based on a threshold, a high similarity class, a medium similarity You may comprise so that it may classify | categorize into the user group which belongs to the similarity class of three classifications, a class and a low degree similarity class.

  For example, when cosine similarity is used as the similarity, 0.6 and 0.4 are set as thresholds, and users whose calculated similarity is in the range of 0.6 or more and 1.0 or less are set to a high degree of similarity. A user group having a class and similarity of 0.4 or more and less than 0.6 as a medium similarity class, and a user having a similarity of 0 or more and less than 0.4 as a low similarity class Is divided.

  In addition, the user group classification unit 354 includes a sorting unit that sorts users in the user group belonging to the high similarity class, the intermediate similarity class, and the low similarity class based on a predetermined rule regarding the similarity. Provided. The sorting unit constitutes the sorting means of the present invention. That is, the sorting unit sorts the users in the user group belonging to the high degree similarity class, the medium degree similarity class, and the low degree similarity class based on a predetermined rule related to the similarity, and classifies them according to the threshold value.

  For example, the sorting unit sorts users belonging to the high similarity class user group and users belonging to the medium similarity class user group in descending order, while sorting users belonging to the low similarity class user group. Sort in ascending order.

  In addition, the user group sorting unit 354 extracts an arbitrary number of users from the classified user groups. For example, the top several people (for example, 10 people) are extracted from the user group classified into each class. However, the user to be extracted may be performed according to different standards for each class. For example, in the high similarity class, the top 10 users are extracted from the users sorted in descending order, and in the medium similarity class, 10 users belonging to the middle range are extracted from the users sorted in descending order. In the low degree similarity class, users are extracted such that the lower 10 users are extracted from the users sorted in ascending order.

(Unexpectedness category extraction unit)
The unexpected category extracting unit 355 constitutes the unexpected category extracting means of the present invention. The unexpected category extraction unit 355 extracts a category that is unexpected to the requesting user by using books purchased or browsed by each user constituting the high-degree similarity class user group. Specifically, the book purchased or browsed by each user constituting the high-degree similarity class user group and the book already purchased or browsed by the requesting user are compared for each category. Then, a category in which a predetermined relationship is established between the number of purchases or the number of browsing times for each category is extracted as an unexpected category. The predetermined relationship here means, for example, a relationship in which a value obtained by quantifying the correlation between the number of purchases or the number of browsing for each category satisfies a predetermined condition.

  For example, a predetermined relationship in each category, that is, a relationship in which a value obtained by quantifying the correlation between the number of purchases or the number of browsing times for each category satisfies a predetermined condition is a user group belonging to a high degree of similarity class. A predetermined difference such as 5 or 10 is obtained by subtracting the number of purchases that the book has already purchased or the number of times of browsing viewed by the requesting user from the number of purchases or the number of times viewed that each user constituting the book has purchased. Categories that are greater than or equal to the value are extracted as unexpected categories.

  In addition, the unexpected category extraction unit 355 executes processing for extracting, for each category extracted as the unexpected category, a book purchased by the user belonging to the user group of the high degree of similarity class. This extracted book is an item that becomes a source of recommendation information on a book to be recommended to the requesting user later.

(Standard frequency calculation part)
The standard degree calculation unit 356 constitutes a standard degree calculation unit of the present invention. This standard degree calculation unit 356 is purchased or viewed by users of a group of users belonging to the intermediate similarity class and the low similarity class from books included in the category extracted by the unexpected category extraction unit 355. The standard degree is calculated based on the number of books purchased or viewed. The standardity here means a value obtained by quantifying an index as a degree of recognition that a book as each item is known to the user.

  For example, in the case of a book belonging to a certain category, it can be said that an introductory book for beginners has a relatively high number of users who recognize the book and a high standard. On the other hand, in books belonging to a certain category, specialized books with extremely high expertise can be said to have a low standardity because the number of users who have such specialized books is relatively small. That is, the standardity here is an index representing the degree of recognition.

(Item extraction unit)
The item extraction unit 357 constitutes item extraction means of the present invention. This item extraction unit 357 is an item having a predetermined standardity from among books as items included in the category extracted by the unexpected category extraction unit 355 based on the standardity calculated by the standardity calculation unit 356. As a book.

  For example, the item extraction unit 357 extracts only books with the highest standardity, or extracts the top five or tenth books from the books with the highest standardity.

(Information provision department)
The information providing unit 358 constitutes information providing means of the present invention. Information relating to the book extracted by the item extraction unit 357 is provided to the requesting user as recommendation information. When providing recommended information to a requesting user, the information providing unit 358 cooperates with the communication control unit 320 to transmit the recommended information to the communication terminal device 10 possessed by the requesting user.

  The recommended information to be provided includes, for example, a book name, text data, a display image, a category to which the book belongs, and the like registered in a category information DB 343 shown in FIG.

[3] Specific Example of Data Registered in Database Next, the details of each database constituting the recording unit 340 included in the information providing server device 30 will be specifically described with reference to FIGS. 3 to 5 described above. explain. 3 is a diagram showing an example of data registered in the user information DB 341, FIG. 4 is a diagram showing an example of data registered in the content information DB 342, and FIG. 5 is registered in the category information DB 343. It is a figure which shows an example of performed data.

(User information DB)
The user information DB 341 is a DB for storing data related to basic information of users who use the information providing system 1 of the present embodiment. In the user information DB 341, as shown in FIG.
(1) User ID for identifying each user,
(2) A password as login information necessary for each user to log in to the information providing system 1 and (3) a book list as a list of items already purchased or viewed by each user are registered in association with each other. Is done.

  In the case illustrated in FIG. 3, the user information DB 341 includes, for example, a user ID of a user specified by “U00001” and a password “ABC1234” required when the user ID logs into the information providing system 1. , A list of books purchased by the user so far, “book 1, book 2, book 3, book 4,...” Is associated and registered. In addition, basic information about the user specified by each user ID, for example, the user's name, the user's gender, the user's date of birth, the user's address, the user's occupation, and the like may be registered.

(Content information DB)
The content information DB 342 is a DB in which various information related to books as items constituting the content is stored as data for each book. In the content information DB 342 shown in FIG.
(1) Book ID for identifying a book as each item,
(2) The book name of each book identified by each book ID,
(3) Represents the contents of each book, book text data registered by text data,
(4) Image data of the cover of each book, and (5) Category ID of the category to which each book belongs,
Are registered in association with each other.

  In the content information DB 342 illustrated in FIG. 4, for example, a book ID “B0001”, a book name “book 1”, “text data 1” representing the book text, and “image 1” which is an image of the cover of the book. And a category ID “C0001” for specifying the category to which the book belongs are registered in association with each other.

(Category information DB)
The category information DB 343 is a DB in which information about a category to which a book as an item belongs is registered for each category. In the category information DB 343 illustrated in FIG.
(1) A category ID for specifying a category and (2) a category specified by each category ID are registered in association with each other.

  For example, in the category information DB 343 illustrated in FIG. 5, the category ID “C0001” and the category “SF” specified by the category ID are registered in association with each other.

[4] Specific Content of Processing Next, the content of processing executed by the data processing unit 350 will be specifically described with reference to FIGS. FIG. 6 is a schematic diagram showing an overview of the overall flow of the system for providing recommendation information to a requesting user, and FIG. 7 is an explanatory diagram for explaining the principle of extracting users with similar preferences. FIG. 8 is an explanatory diagram for explaining a book extraction function that is included in a predetermined category and gives a user a surprise, and FIG. 9 is an explanatory diagram for explaining a process of calculating a standard degree. .

  In the information providing server device 30 of the present embodiment, first, the counting unit 352 counts the number of books purchased by each user so far for each category between the requesting user and other users.

  In the example shown in FIG. 6, the requesting user who receives the recommendation purchased 25 books that belong to the category of SF, 1 who purchased the books that belong to the mystery category, and purchased books that belong to the horror category. The number is twelve. Similarly, IT is 1, physical is 0, business is 2, and qualification is 5. In contrast, the number of books per category that one of the other users has purchased so far is 30 for SF, 10 for mystery, 10 for horror, 5 for IT, 0 for physics, 0 for business Is 6.

  Next, the similarity calculation unit 353 calculates the similarity between the requesting user and another user. In the similarity calculation method, first, a preference vector is created based on the categories of books that each user has purchased so far. The similarity is calculated using the cosine similarity based on the created preference vector. However, the method of calculating the similarity is not particularly limited, and the similarity may be calculated using, for example, the Euclidean distance.

  Based on the calculated similarity, the user group classifying unit 354 has a high similarity user group with a high degree of similarity, a medium intermediate similarity user group, and a low low similarity degree. Divide into user groups. In FIG. 6, the area shown in FIG. 6 shows a user group with a high degree of similarity and a high degree of similarity, and the explanatory text explained by (2) in FIG. This is a process of dividing the user group into a user group having a low degree and a user group having a low degree of similarity.

  When classifying each user into a user group of each class, the sorting unit sorts, for each user group, users belonging to the high similarity class and users belonging to the intermediate similarity class in descending order while low Sort users belonging to the similarity class in ascending order. Further, the data processing unit 350 extracts an arbitrary number of users from the classified user groups. For example, the top 10 users are extracted from the group of users classified into each class. However, the number of people to be extracted may be other numbers as necessary.

  Then, the unexpected category extraction unit 355 determines, for each category, between the ten users extracted from other users and the requesting user for the user group having a high degree of similarity among the divided user groups. Then, the difference between the purchased books is obtained, and the category having a large difference is extracted. In the example shown in FIG. 6, when the difference in the number of purchased books is 5 or more, the difference is assumed to be large. As a result, categories belonging to SF and categories belonging to the mystery are extracted.

  Next, as shown in (1) of FIG. 6, the requested user who receives the recommendation extracts a book that is an unpurchased item. At that time, as shown in the explanatory text of (3) in FIG. 6, the filtering unit purchases by any of the users belonging to the moderate similarity user group and the users belonging to the low similarity user group. A process of eliminating books that have never been performed is executed. Conversely, books that have been purchased by both user groups are left without being excluded.

  Next, as shown in the explanatory text of (4) in FIG. 6, the standard degree calculation unit 356 executes a pointing process on the extracted book. Specifically, for a book that has been purchased by a user who belongs to the user group of the intermediate class and a user who belongs to the user group of the low class class, a book that has not been purchased has been added 1 points. For, the point is 0. Then, the standard degree calculation unit 356 defines the result of the added points as a standard degree, and as shown in the explanatory text of (5) in FIG. Sort.

  Next, the item extraction unit 357 extracts books included in the top three from the sorted books. However, the item extraction unit 357 can extract only one book having the highest standard degree, or can extract books included in other ranks.

  Then, the information providing unit 358 recommends the book extracted by the item extracting unit 357 as recommended information to the requesting user.

  Regarding the processing described above, the processing of the counting unit 352, the similarity calculating unit 353, and the user classifying unit 354 will be described in more detail with reference to FIG.

  In the example shown in FIG. 7, user A indicates a requesting user who receives a recommendation, and user B to user F indicate other users. In the example illustrated in FIG. 7, the number of books purchased by the user A for each category is 20 for category A, 15 for category B, 10 for category C, 5 for category D, and 0 for category E.

  The number of books purchased for each category purchased by user B, one of the other users, is 23 for category A, 14 for category B, 10 for category C, 20 for category D, and 5 for category E. Similarly, with respect to the number of books purchased for each category purchased for user F from user C who is another user, the number of books as shown in FIG. 7 is purchased for each category.

  In FIG. 7, the right column of the figure shows the similarity of each user. In FIG. 7, the similarity is calculated based on the cosine similarity, and the similarity between the user B and the user F is 0.898553136, 0.1654545563, 0.227333666, 0.561029499, and 0.568008578, respectively. .

  Furthermore, the characters “high”, “medium”, and “low” shown on the right side of FIG. 7 represent the degree of similarity, respectively. In FIG. 7, user B belongs to a user group of a high degree of similarity class, user C and user D belong to a user group of a low degree of similarity class, and user E and user F belong to a medium similarity class. It belongs to the user group. In FIG. 7, users belonging to the range where the similarity is 0.6 or more and 1.0 or less are assigned to the high similarity class, and those belonging to the range where the similarity is 0.4 or more and less than 0.6 A user group is classified as a low degree similarity class, with users belonging to a range of degree similarity class and similarity between 0 and less than 0.4.

  Next, the process of the unexpected category extraction unit 355 will be described with reference to FIG.

  The unexpected category extracting unit 355 requests the number of books purchased in each category purchased by the requesting user and the number of books purchased in each category purchased by the users belonging to the high-degree similarity class user group. Items that give the user a surprise, that is, books are extracted.

Among users having a high degree of similarity, while there are categories where some users do not purchase books, other users may purchase many books belonging to the category. A high degree of similarity indicates that the users have similar preferences. However, when a book in a category that one user has rarely purchased is recommended, the user feels surprised with the recommended book. The unexpected category extraction unit 355 is a configuration unit for executing such processing with the user's psychology.

  Now, a category in which the difference in the number of purchases between the requesting user and the user B is 5 or more is regarded as a surprising category. In FIG. 8, categories whose purchase number difference is 5 or more are category D and category E. That is, the user A who is the requesting user has purchased only 5 books of the category D, whereas the user B has also purchased 20 books of the category D, and the difference is 15. The user A who is the requesting user has not purchased any category E books, whereas the user B has also purchased 5 category E books, and the difference is 5. For other categories A to C, the difference in the number of purchases is less than 5.

  Next, the function of the filtering unit described above will be described in detail with reference to FIG.

  In the information providing system 1 of the present embodiment, as described above, one of the purposes is to provide the requesting user with information related to books with a high standardity. However, books that have never been purchased or viewed by users belonging to the user group of the medium similarity class or users belonging to the user group of the low degree of similarity class are not widely recognized in the world. There is a risk that the standardity is very low. When information about a book with such a low standard is provided to a requesting user, the requesting user may be given a sudden image and may not be interested at all. The filtering unit is a component provided to avoid such a situation.

  “◯” in FIG. 9 indicates that there is a purchase record, and “×” indicates that there is no purchase record. In FIG. 9, “books” to “books” are books that have been purchased by both users belonging to the medium similarity class user group and users belonging to the low similarity class user group. It can be seen that it is.

  On the other hand, “Book” has never been purchased by a user belonging to the moderate similarity class user group and has been purchased by a user belonging to the low similarity class user group. “To Book” has been purchased by a user belonging to the user group of the medium similarity class, but has not been purchased by a user belonging to the user group of the low similarity class. “Book” has never been purchased by both a user belonging to the user group of the medium similarity class and a user belonging to the user group of the low similarity class.

  Therefore, “books” to “books” shown in the frame remain as recommended books, but “books” to “books” shown outside the frame do not remain as recommended books. The book is not recommended.

[5] Operation of Information Providing System of the Present Embodiment Next, a process in which the information providing system 1 of the present embodiment provides recommended information to the requesting user will be described with reference to FIG. FIG. 10 is a flowchart showing processing executed by the information providing server device 30.

  In this operation, it is assumed that initialization processing for the recommendation engine has already been executed. It is assumed that necessary data is already registered in the user information DB 341, the content information DB 342, and the category information DB 343.

  First, the information providing server device 30 determines whether or not the communication control unit 320 has received a signal indicating that a user login operation has been performed from the communication terminal device 10 (step S101).

  Next, the counting unit 352 counts the number of purchases or the number of browsing of items such as books purchased or browsed by each user for each category to which the book belongs (step S102). In this step, when a book is taken as an example as an item, the categories are, for example, the fields of SF, mystery, IT, and business. The counting unit 352 counts each user such that the category is 25 for SF and the category is 1 for mystery.

  Next, the similarity calculation unit 353 calculates the preference similarity between the requesting user and other users for each other user. (Step S103). The calculation method of the similarity is not particularly limited, and can be calculated by various methods such as calculation using a cosine similarity calculation method or calculation using a Euclidean distance. In the example shown in FIG. 7, the similarity calculated by the method of calculating the cosine similarity is shown.

  Next, the user group classifying unit 354 classifies the user into a user group of a high degree of similarity class, a user group of a medium similarity, and a user group of a low degree of similarity based on the calculated similarity (step S104). The user group classification unit 354 performs classification by setting a predetermined numerical range for the similarity for each class of user groups. For example, the user group classification unit 354 indicates that users who belong to a range whose similarity is 0.6 or more and 1.0 or less are users who belong to a high degree of similarity class and whose similarity is 0.4 or more and less than 0.6. Are classified into a medium similarity class, and users belonging to a range of similarity between 0 and less than 0.4 are classified as low similarity classes.

  In step S104, the sorting unit also executes a process of sorting the users belonging to each user group in ascending order or descending order according to the magnitude of similarity. For example, users belonging to the high degree of similarity class and users belonging to the medium degree of similarity class are sorted in descending order, while users belonging to the low degree of similarity class are sorted in ascending order.

  In step S104, a predetermined number of users are extracted from the user group of each class. For example, the number of users belonging to the high-degree similarity class user group sorted in descending order is five or ten users from the top. In the user group of the high degree similarity class, when the users are sorted in descending order and users such as 5 or 10 are extracted from the top, the extracted user is the degree of similarity with the requesting user. The characteristic of being high can be made remarkable.

  The users belonging to the user group of the medium similarity class sorted in descending order are the same as the users belonging to the user group of the high similarity class. However, in order to make the characteristic that the degree of similarity is medium among the users belonging to the user group of the medium similarity class, the rank is in the middle range among the users sorted in descending order. It is good to extract a user.

  On the other hand, the number of users belonging to the low-degree similarity class user group sorted in ascending order is five or ten users from the top. That is, 5 or 10 people are extracted in order from the user with the lowest similarity. When the users are sorted in ascending order in the user group of the low degree similarity class and the users such as 5 or 10 from the top are extracted, the extracted users are the degree of similarity with the requesting user. The characteristic of low is remarkable.

  Next, the unexpected category extracting unit 355 selects an unexpected category based on the number of books purchased for each category by the requesting user and the number of books purchased for each category by the users belonging to the user group of the high degree of similarity class. Extract (step S105).

  The unexpected category extraction unit 355 includes a category having a certain rule between the number of books purchased for each category by the requesting user and the number of books purchased for each category by the users belonging to the high-degree similarity user group. Are extracted as unexpected categories. For example, if there is a difference of a certain value or more between the number of books purchased by the requesting user and the number of books purchased by the users belonging to the user group of the high degree of similarity class, the category is set as an unexpected category. Extract. In the example illustrated in FIG. 8, the category D and the category E in which the difference in the number of purchases is 5 or more are extracted as the unexpected categories.

  Further, in step S105 for extracting the unexpected category, a process of extracting a book purchased by the user belonging to the user group of the high similarity class is executed for each category extracted as the unexpected category.

  Next, the filtering unit performs filtering on the book that is included in the category extracted as the unexpected category and purchased by the user belonging to the user group of the high degree of similarity class (step S106).

  The filtering unit excludes, from among the extracted books, books that have never been purchased by either users belonging to the medium similarity class user group or users belonging to the low similarity class user group. Execute the process. Conversely, books that have been purchased by both users belonging to the medium similarity class user group and users belonging to the low similarity class user group are left as they are.

  Next, the standardity calculation unit 356 calculates the standardity of books included in the category extracted by the unexpected category extraction unit 355 (step S107).

  The standardity calculation unit 356 is a book purchased or viewed by users of a group of users belonging to the intermediate similarity class and the low similarity class from among the books included in the category extracted by the unexpected category extraction unit 355. The standard degree is calculated based on the number of purchases or the number of browsing. Specifically, for a book that has been purchased by a user who belongs to the user group of the intermediate class and a user who belongs to the user group of the low class class, a book that has not been purchased has been added 1 points. For, the point is 0. Then, the standard degree calculation unit 356 defines the result of the added points as a standard degree, and sorts the books in the order of the highest points, that is, the highest standard degree.

  Next, the item extraction unit 357 extracts a book with a predetermined standard frequency (step S108). For example, the item extraction unit 357 extracts only books with the highest standardity, or extracts the top five or tenth books from the books with the highest standardity.

  Thereafter, the information providing unit 358 transmits information regarding the book extracted by the item extracting unit 357 to the requesting user (step S109), and the operation is terminated.

  The information providing server device 30 constituting the information providing system 1 in the present embodiment executes the process of providing recommended information to the requesting user with the above-described configuration, so that the requesting user feels unexpectedness, Provide acceptable information.

[6] Modification [6.1] Modification 1
The information providing system 1 of the present embodiment employs a field to which an item belongs as a category. However, the category is not limited to the field to which the item belongs.

  In the information providing system 1 according to the first modification, an item author, for example, a book author, a music lyricist, a composer, or the like may be employed as a category. When such an author is adopted as a category, the requesting user has little knowledge, but it is possible to provide the requesting user with information about the author's books that are intriguing and the music of the songwriter and composer.

[6.2] Modification 2
In the information providing system 1 according to the present embodiment, the description has been made mainly on the purchased books. However, instead of the purchased books, the information on the browsed books can be accumulated and the process can be executed.

  In this case, for example, when the last page of the book is opened by the book browsing application, information indicating that the last page has been opened is transmitted from the communication terminal device 10. It transmits to the information providing server device 30. Since it is presumed that the user has read the book to the last page, the information indicating that it has been opened up to the last page increases the probability that the requesting user who receives the recommendation information feels useful.

[6.3] Modification 3
In the similar user extraction function according to the present embodiment, the process may be executed by collecting users having similar similarities together.

  For example, in a group of users in the intermediate similarity class, users who belong to each range of cosine similarity of 0.4 units and 0.5 units are collected together, and purchased books are handled for the similarity of each range. May be. In the third modification, a good result can be obtained even when the amount of data is small.

[6.4] Modification 4
In the similar user extraction function in the present embodiment, the process may be executed by excluding a user having a very low similarity with the requesting user. For example, a user whose cosine similarity with the requesting user is less than 0.2 is excluded from processing executed by the information providing server device 30 in the information providing system 1.

  In this modified example 4, since users with extremely low similarity are excluded, the unexpectedness of the information provided to the requesting user is certainly reduced. However, by excluding users with extremely low similarity, it is possible to provide information that is easily received by the requesting user and useful for the requesting user.

[6.5] Modification 5
In the extraction function of the unexpected category extraction unit 355 in this embodiment, instead of the logic of eliminating all categories in which the difference in the number of purchases for each category is smaller than a predetermined value, the higher-order categories with the large difference in the number of purchases are displayed. The logic of adopting may be adopted. For example, the logic is to employ the top three categories with the largest difference in the number of books purchased in the category.

[6.6] Modification 6
In the function provided by the standard degree calculation unit 356 of the present embodiment, the user group of each class is weighted with respect to the standard degree in the user group of the intermediate similarity class and the user group of the low degree similarity class. The standardity may be calculated.

  For example, when calculating the standard degree of the user group of the medium similarity class, a process of increasing the weight of the standard degree of the user group of the medium similarity class is performed. In this case, unexpectedness is reduced for the requesting user, but it is possible to provide information that is easily received and useful. On the other hand, when calculating the standardity of the user group of the low degree similarity class, a process of increasing the weight of the standard degree of the user group of the low degree similarity class is performed. In this case, unexpectedness can be increased for the requesting user.

[6.7] Modification 7
In the present embodiment, the DBs 341 to 343 are provided in the information providing server device 30, and these DBs are managed and controlled only by the information providing server device 30. However, a computer for managing the DBs 341 to 343 is used. Each system may be provided separately and managed and operated individually.

DESCRIPTION OF SYMBOLS 1 ... Information provision system 10 ... Communication terminal device 20 ... Network 30 ... Information provision server device 310 ... Server management control part 320 ... Communication control part 330 ... ROM / RAM
340 ... Recording unit 341 ... User information DB
342 ... Content information DB
343 ... Category information DB
350 ... Data processing unit 351 ... DB management unit 352 ... Counting unit 353 ... Similarity calculation unit 354 ... User group classification unit 355 ... Unexpectedness category extraction unit 356 ... Standard degree calculation unit 357 ... Item extraction unit 358 ... Information providing unit

Claims (8)

An information providing server device connected to a communication terminal device possessed by a user via a network and providing the communication terminal device with information related to an item that constitutes one type of predetermined content to the user as recommendation information,
A counting means for counting the number of purchases or the number of browsing of the item purchased or viewed by each user for each category to which the item belongs;
A similarity calculation means for calculating a similarity regarding purchase or browsing preferences for the content among users based on the number of purchases or the number of browsing;
Based on the calculated similarity between the users, other users with respect to the requesting user requesting the recommendation information are classified into a high similarity class, a medium similarity class, and a low similarity class based on a threshold. A user group dividing means for dividing into user groups belonging to the three classification similarity classes;
The item purchased by or viewed by each user constituting the user group of the high degree of similarity class and the item already purchased or viewed by the requesting user are compared for each category, and the number of purchases or browsing for each category. A surprising category extracting means for extracting, as a surprising category, a category in which a relationship in which a value obtained by quantifying the correlation between the number of purchases or the number of times of browsing for each category satisfies a predetermined condition is established;
Number of purchases or browsing of items purchased or viewed by users of the user group belonging to the intermediate similarity class and the low similarity class from items included in the category extracted by the unexpected category extraction unit Based on the number of times, a standard degree calculating means for obtaining a standard degree that quantifies an index as a degree of recognition of each item,
Based on the standardity calculated by the standardity calculation means, item extraction means for extracting an item having a predetermined standardity from items included in the category extracted by the unexpected category extraction means;
An information provision server apparatus comprising: an information provision stage that provides information relating to the item extracted by the item extraction means to the requesting user as the recommendation information. In the information provision server apparatus of Claim 1,
The user group classifying means sorts the users of the user group belonging to the high degree similarity class, medium degree similarity class, and low degree similarity class based on a predetermined rule relating to the similarity, and classifies them according to the threshold value. An information providing server device in which means is provided for each class. In the information provision server apparatus of Claim 1 or 2,
The predetermined condition in each category is the number of purchases or purchases that the requesting user has already purchased the item from the number of purchases or the number of browsing times that each user constituting the user group belonging to the high degree of similarity class purchased the item. An information providing server device in which a difference obtained by subtracting the number of browsing times is equal to or greater than a predetermined value. In the information provision server apparatus of any one of Claims 1-3,
Items that have never been purchased or viewed by users of the user group belonging to the intermediate similarity class and the low similarity class among the items included in the category extracted by the unexpected category extraction means An information providing server device, further comprising filtering means for eliminating. In the information provision server apparatus of any one of Claims 1-4,
The information providing server device, wherein the standard degree calculating means calculates the standard degree by calculating the number of items purchased or the number of browsed items for each item. In the information provision server apparatus of any one of Claims 1-5,
The item extraction means is an item that the requesting user has not purchased or browsed yet, and an item that has already been purchased or browsed by a user belonging to the user group of the high degree of similarity class is extracted. Information providing server device. In a computer constituting an information providing server device connected to a communication terminal device possessed by a user via a network and providing the communication terminal device with information related to an item constituting one type of predetermined content to the user as recommendation information A program used,
The computer,
A counting means for counting the number of purchases or the number of browsing of the item purchased or viewed by each user for each category to which the item belongs,
Similarity calculation means for calculating a similarity regarding purchase or browsing preferences for the content among users based on the number of purchases or the number of browsing.
Based on the calculated similarity between the users, other users with respect to the requesting user requesting the recommendation information are classified into a high similarity class, a medium similarity class, and a low similarity class based on a threshold. User group dividing means for dividing into user groups belonging to the similarity classes of
The item purchased by or viewed by each user constituting the user group of the high degree of similarity class and the item already purchased or viewed by the requesting user are compared for each category, and the number of purchases or browsing for each category. A surprising category extracting means for extracting, as an unexpected category, a category in which a relationship in which a value obtained by quantifying the correlation between the number of purchases or the number of browsing for each category satisfies a predetermined condition is established,
Number of purchases or browsing of items purchased or viewed by users of the user group belonging to the intermediate similarity class and the low similarity class from items included in the category extracted by the unexpected category extraction unit Based on the number of times, a standard degree calculation means for obtaining a standard degree obtained by quantifying an index as a degree of recognition of each item,
Item extraction means for extracting an item having a predetermined standardity from items included in the category extracted by the unexpected category extraction means based on the standardity calculated by the standardity calculation means, and the item An information provision stage for providing information relating to the item extracted by the extraction means to the requesting user as the recommendation information;
A program characterized by functioning as An information providing method for providing information relating to an item that is connected to a communication terminal device possessed by a user via a network to the user as recommendation information to the communication terminal device,
A counting step of counting the number of purchases or the number of browsing of the item purchased or viewed by each user for each category to which the item belongs;
Based on the number of purchases or the number of times of browsing, a similarity degree calculating step for calculating the degree of similarity regarding purchase or browsing preferences for the content among users;
Based on the calculated similarity between the users, other users with respect to the requesting user requesting the recommendation information are classified into a high similarity class, a medium similarity class, and a low similarity class based on a threshold. A user group dividing step of dividing into user groups belonging to the three classification similarity classes;
The item purchased by or viewed by each user constituting the user group of the high degree of similarity class and the item already purchased or viewed by the requesting user are compared for each category, and the number of purchases or browsing for each category. An unexpected category extraction step of extracting a category in which a relationship in which a value obtained by quantifying the correlation between the number of purchases or the number of browsing for each category is satisfied between predetermined times is established as an unexpected category;
Number of purchases or browsing of items purchased or viewed by users of the user group belonging to the intermediate similarity class and the low similarity class from items included in the category extracted by the unexpected category extraction unit Based on the number of times, a standard degree calculation step for obtaining a standard degree that quantifies an index as a degree of recognition of each item,
Based on the standardity calculated by the standardity calculation means, an item extraction step for extracting items having a predetermined standardity from items included in the category extracted by the unexpected category extraction means;
An information providing method comprising: providing information relating to the item extracted by the item extracting means to the requesting user as the recommended information.

Images