JP5235227B2 - Log processing apparatus and operation method thereof - Google Patents

Description

  The present invention relates to a log processing apparatus and an operation method thereof.

  Collaborative filtering is a technology that, in a broad sense, accumulates preference information of many users and recommends content to relevant users using information on other users who have similar preferences to a certain user (user).

  This technology can be classified into an inter-user type and an item (also called content) type. The inter-user type is a technique that first finds a user (referred to as a similar user) whose preference pattern is similar to the user receiving the recommendation, and sets an item group preferred by the similar user as a recommended candidate. For implementation, the degree of similarity between users is represented by the correlation coefficient (Pearson correlation, rank correlation, etc.) of the evaluation given to the same item, and users with high similarity are extracted for preference prediction. The evaluation value for the item is weighted with the similarity to the user, and the weighted average value of the evaluation values is used as a prediction. Recommendations are made as recommended items from those with large predicted values. Since there is a restriction on the size of the display screen, an item having a small predicted value is deleted, or the upper 3 to 10 items are displayed.

  Also, in the item-to-item type, it is assumed that items that receive similar evaluations by various users are similar, and the assumption that users are interested in similar items of the items they are interested in is the closest to the user. Recommend items that are similar to items in your usage history. As an implementation, the similarity between items is measured based on the co-occurrence of the user of the item, and the similar item of the item in the latest usage history of the user is recommended.

  In this case, assuming the application within a closed range for one service, enter the history information of the target item that took action history such as user browsing, click, reservation, purchase, viewing, evaluation, etc. and the item metadata It is assumed that.

  Details of these techniques are described in Non-Patent Document 1.

Toshihiro Kamijo, “Algorithm of Recommended System (2),” Journal of Japanese Society for Artificial Intelligence Vol.23, No.1, pp.89103, January 2008

  When using the user-to-user recommendation method, the weighted average value of evaluation values for a certain item is used, so when the item is newly registered as content, etc., at a stage where there are few evaluation logs for that item Since the evaluation value of an item other than that item is considered to have a larger value than that item, there is a problem that it is not displayed as a recommendation. This means that recommendation across services does not function correctly when the data amount of item IDs, the frequency of registration / deletion, the quality of metadata, etc. differ between services.

When using the item-to-item type recommendation method, the similarity between items is measured by the co-occurrence of the user who uses the item. In a small number of stages, there is a problem that the degree of similarity to other items is not displayed as a recommendation because it is considered to have a larger value than that item. This means that the recommendation across services does not function correctly when the data amount of the user ID, the frequency of new registration / unsubscription, the usage frequency, etc. differ between services.
These issues are known as cold start problems.

  In addition, even when the recommendation is performed by integrating the ID system, it is difficult to recommend with the same accuracy as the recommendation within the range closed to one service because the quality of the ID system and log data varies. .

  Furthermore, assuming an actual service, the database and the number of IDs are limited, and it is necessary to reuse IDs used in the past as they continue to be used. In particular, if the ID is reused without processing, past data and current data are mixed, and the accuracy of the output of the recommendation processing is lowered.

  The present invention has been made in view of the above problems, and an object of the present invention is to provide a log processing apparatus and an operation thereof for eliminating inconvenience caused by a difference between a user ID and a content ID in a plurality of services. It is to provide a method.

  In order to solve the above-described problem, the first aspect of the present invention is a log that processes a log accumulated to recommend content to a destination user when content can be transmitted in each of a plurality of services. A processing device, a service ID of the service indicating that any content of any service has been transmitted for any user, a content ID of the content at the service, and related to the content A log storage unit that stores logs including content-related information, a user ID of the user in the service, and user-related information related to the user, and a record generated for each log of the log storage unit are stored. Log record storage unit for storing a user record including a plurality of user IDs indicating the same user in a plurality of services Integrated user ID database, integrated content ID database storing content records including a plurality of content IDs indicating the same content in a plurality of services, and a plurality of logs having the same user related information and different service IDs A user ID processing unit that reads a user ID from each of the user ID and stores a user record including the user ID and the integrated user ID in the integrated user ID database, and a plurality of logs having the same content related information and different service IDs. A content ID processing unit that reads out content IDs from each of them, stores a content record including the content ID and the integrated content ID in the integrated content ID database, and a log for the log storage unit. The integrated user ID searched from the integrated user ID database for the record containing the user ID and read from the record, and the integrated content ID database searched for the record containing the content ID in the log and read from the record And a log record generation unit that stores a record including the integrated content ID and the service ID in the log in the log record storage unit. The second aspect of the present invention is an operation method of a log processing apparatus that performs processing of a log accumulated to recommend content to a destination user when content can be transmitted in each of a plurality of services, The log processing device relates to the service ID of the service indicating that any content of any service has been transmitted for any user, the content ID of the content in the service, and the content A log storage unit that stores logs including content-related information, a user ID of the user in the service, and user-related information related to the user, and a record generated for each log of the log storage unit are stored. Log records storage unit for storing multiple user IDs indicating the same user in multiple services. And an integrated content ID database in which content records including a plurality of content IDs indicating the same content in a plurality of services are stored, and the operation method includes a user ID of the log processing device The processing unit reads a user ID from each of a plurality of logs having the same user-related information and different service IDs, and stores a user record including the user ID and the integrated user ID in the integrated user ID database, and the log A content ID processing unit of the processing device reads the content ID from each of a plurality of logs having the same content-related information and different service IDs, and stores the content record including the content ID and the integrated content ID in the integrated content The log record generation unit of the log processing device searches for a record including the user ID in the log from the integrated user ID database for each log of the log storage unit and reads out the record from the record A record including the integrated user ID and a record including the content ID in the log from the integrated content ID database and reading the integrated content ID read from the record and the service ID in the log in the log record storage unit. A log processing apparatus operating method characterized by storing the information is used as a solving means.

  ADVANTAGE OF THE INVENTION According to this invention, the technique for eliminating the inconvenience resulting from the difference in user ID and content ID in a some service can be provided.

It is a figure which shows the structure of the recommendation server 1 using the log processing apparatus which concerns on this Embodiment. 3 is a diagram illustrating an example of contents of a log storage unit 101. FIG. 5 is a diagram illustrating an example of contents of a service number storage unit 104. FIG. 5 is a diagram illustrating an example of contents of a service number storage unit 105. FIG. It is a figure which shows an example of the content of the integrated user ID database. It is a figure which shows an example of the content of the integrated content ID database. 3 is a diagram illustrating an example of contents of a log record storage unit 110. FIG. 5 is a flowchart showing an operation of a log record deletion unit 112. 5 is a flowchart showing an operation of a log deletion unit 114.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings.

  FIG. 1 is a diagram showing a configuration of a recommendation server 1 using the log processing apparatus according to the present embodiment.

  The recommendation server 1 is a device that recommends content to a destination user when content can be transmitted in each of a plurality of services.

  The recommendation server 1 includes (1) a log accumulation unit 101 that accumulates logs as a history indicating that content is transmitted in a plurality of services, and (2) a preprocessing using the log accumulation unit 101 for content recommendation. Processing using the preprocessing units 102 and 103 to be performed, (3) service number storage units 104 and 105 in which data obtained by the preprocessing is stored, and (4) log storage unit 101 and service number storage unit 104 A user ID processing unit 106 that performs processing, (5) a content ID processing unit 107 that performs processing using the log storage unit 101 and the service number storage unit 105, and (6) a record generated by the user ID processing unit 106. Integrated user ID database 108 stored, and (7) integrated content storing records generated by the user ID processing unit 107 D database 109, (8) log storage unit 101, integrated user ID database 108, log record storage unit 110 in which records generated using integrated content ID database 109 are stored, and (9) generate the records A log record generation unit 111; (10) a log record deletion unit 112 that deletes a record from the log record storage unit 110; (11) a content recommendation unit 113 that recommends content using the log record storage unit 110; 12) a log deletion unit 114 that deletes a log from the log storage unit 101;

FIG. 2 is a diagram illustrating an example of the contents of the log storage unit 101.
The log storage unit 101 stores a plurality of logs. Each log indicates that any content of any service has been transmitted for any user. The transmission destination is, for example, a terminal (such as a personal computer or a portable terminal) used by the user.

  Each log includes (1) identification information indicating the corresponding service (hereinafter referred to as “service ID”), and (2) identification information indicating the content of the service (hereinafter referred to as “service ID”). (Content ID)), (3) information related to the content (hereinafter referred to as “content-related information”), (4) identification information indicating the user at the service (hereinafter referred to as service) (Referred to as “user ID”) and (5) information related to the user (hereinafter referred to as user-related information).

  The content related information is, for example, a title of the corresponding content, a synopsis, and the like. Here, “content-related information” is used as a name indicating each element such as a title and a generic name of each element. Note that each content-related information is not necessarily included in all service logs, but for convenience, such information is referred to as content-related information.

  The user related information is, for example, the name, telephone number, etc. of the corresponding user. Here, “user-related information” is used as a name indicating each element such as a name and a generic name of each element. Note that each user-related information is not necessarily included in all service logs, but for convenience, such information is referred to as user-related information.

  Further, although not shown, each log includes (1) a time when the user clicked (click time) when transmitting the content, and (2) a time when the user made a reservation when transmitting the content (reserved time). ), (3) the time when the user performed a purchase operation (purchase time), (4) the time when the user viewed the content (viewing time), and (5) the user evaluated the content One or more times (evaluation times) are included. Further, the log including the evaluation time may include a score given in the evaluation.

  For each type of user related information, the pre-processing unit 102 counts the number of service IDs (referred to as the number of services) in the log having the user related information belonging to the type, and (1) the user A record having information indicating the type of related information and (2) the number of services is generated and stored in the service number storage unit 104.

FIG. 3 is a diagram illustrating an example of the contents of the service number storage unit 104.
FIG. 3 shows (1) a record including information “name” indicating the type of user-related information, (2) a service number “3”, (1) “phone number” and (2) service number “3”. A record including

  Returning to FIG. 1, the preprocessing unit 103 counts the number of service IDs (referred to as the number of services) in the log having content related information belonging to the type for each type of content related information in the log storage unit 101. A record having (1) information indicating the type of the content-related information and (2) the number of services is generated and stored in the service number storage unit 105.

FIG. 4 is a diagram illustrating an example of the contents of the service number storage unit 105.
FIG. 4 shows (1) information “title” indicating the type of content-related information, and (2) a record including the number of services “3”, (1) “Synopsis” and (2) the number of services “3”. Examples include records.

  Returning to FIG. 1, the user ID processing unit 106 reads out the user ID from each of a plurality of logs having the same user-related information in the log storage unit 101, and (1) the user ID and (2) the corresponding user ID. A record including an integrated user ID indicating a user is generated and stored in the integrated user ID database 108.

  At this time, the user ID processing unit 106 first reads information indicating the type of user-related information from the record having the largest number of services in the service number storage unit 104, and (1) includes the same user-related information belonging to the type. (2) Search a plurality of logs including different service IDs, read out the user ID from each log, (1) each user ID, and (2) an integrated user ID indicating the corresponding user, Are generated and stored in the integrated user ID database 108.

  Thereafter, the user ID processing unit 106 performs the same operation in descending order for the second and subsequent service numbers in the service number storage unit 104.

FIG. 5 is a diagram showing an example of the contents of the integrated user ID database 108.
FIG. 5 exemplifies a record including (1) three user IDs “9648”, “U36128”, “User6198”, and (2) an integrated user ID “T00001”.

  Returning to FIG. 1, the content ID processing unit 107 reads the content ID from each of a plurality of logs having the same content-related information in the log storage unit 101, and (1) the content ID and (2) the corresponding ID. A record including the integrated content ID indicating the content is generated and stored in the integrated content ID database 109.

  At this time, the content ID processing unit 107 first reads information indicating the type of content related information from the record having the largest number of services in the service number storage unit 105, and (1) includes the same content related information belonging to the type. (2) Search a plurality of logs including different service IDs, read out the content ID from each log, (1) each content ID, and (2) an integrated content ID indicating the corresponding content, Are generated and stored in the integrated content ID database 109.

  Thereafter, the content ID processing unit 107 performs the same operation in descending order for the second and subsequent service numbers in the service number storage unit 105.

FIG. 6 is a diagram showing an example of the contents of the integrated content ID database 109.
FIG. 6 exemplifies a record including (1) three content IDs “100056”, “C68913”, “Content8916”, and (2) an integrated content ID “C00001”.

  Returning to FIG. 1, the log record generation unit 111 performs (1) all the click time, reservation time, purchase time, viewing time, and evaluation time in the log for each log in the log storage unit 101, and (2) the current Time (hereinafter referred to as log generation time), (3) an integrated user ID retrieved from the integrated user ID database 108 by searching for a record including the user ID in the log, and (4) an integrated content ID database 109, a record including the content ID in the log is retrieved and a record including the integrated content ID read from the record and (4) the service ID read from the log is generated, and the log record storage unit 110 generates the record. Remember.

  Note that the log record generation unit 111 includes a score read from the corresponding log in the record including the evaluation time. On the other hand, the log record generation unit 111 includes a score “1” in a record that does not include an evaluation time.

FIG. 7 is a diagram illustrating an example of the contents of the log record storage unit 110.
7 shows (1) click time “2011/06/23; 23: 12: 23”, evaluation time “2011/06/29; 23: 45: 33”, and (2) log generation time “2011/06 / 28; 00: 10: 00 ”, (3) integrated user ID“ T12541 ”, (4) integrated content ID“ C15244 ”, (5) service ID“ SV1 ”, and (6) a record including a score Is illustrated.

FIG. 8 is a flowchart showing the operation of the log record deletion unit 112.
The log record deletion unit 112 prevents inconvenience such as recommendation of content described later biased to a certain service by deleting records from the log record storage unit 110.

  First, the log record deletion unit 112 obtains the Gini coefficient GC1 indicating the imbalance of the number of records in the set of records including the service ID for each service ID in the log record storage unit 110, and the Gini coefficient GC1 is determined in advance. It is determined whether it is greater than a threshold value (hereinafter referred to as threshold value GCT1) (S1).

  The Gini coefficient GC1 indicates an imbalance of the number of elements (in this case, the number of records) in each target (in this case, a set of records). The Gini coefficient GC1 is included in the range of 0-1. The closer the Gini coefficient GC1 is to 1, the greater the degree of imbalance, and the closer the Gini coefficient GC1 to 0, the smaller the degree of imbalance. When there is no imbalance, that is, when the number of records is equal to each other, the Gini coefficient GC1 becomes zero. The same applies to the Gini coefficient GC2, which will be described later.

In obtaining the Gini coefficient GC1, the log record deleting unit 112 first calculates the average difference MDF using Equation (1).

  Here, n is the total number of services, xi is the number of records including the service ID of service i, and xj is the number of records including the service ID of service j.

Next, the log record deletion unit 112 calculates the average value μ using Equation (2).

  Here, n is the total number of services, and x1, x2,... Are the numbers of records including the service IDs of the first, second,.

Next, the log record deletion unit 112 calculates the Gini coefficient GC using Expression (3).
GC1 = MDF / (2 × μ) (3)
If the Gini coefficient GC1 is larger than the threshold GCT1 (S1: YES), the log record deletion unit 112 sets a set of records having a larger number of records than a predetermined number (hereinafter referred to as threshold LT1) to the log record storage unit 110. The records are deleted from the set so that the number of records in the set is equal to the threshold LT1 (S3).

  If the Gini coefficient GC1 is less than the threshold value GCT1 (S1: NO), or if the process of step S3 is completed, the log record deletion unit 112 performs log service storage unit 110 for each service ID. The set of records including the service ID is divided. Here, it is assumed that the log record deletion unit 112 includes viewing times (other times, for example, click times) belonging to a time zone having the same time length (for example, 1 hour).

  Next, the log record deletion unit 112 sets the Gini coefficient GC2 indicating the imbalance of the number of records in the record set (hereinafter referred to as a set) for all of the divided record sets regardless of the service ID. Then, it is determined whether or not the Gini coefficient GC2 is larger than a predetermined threshold (hereinafter referred to as threshold GCT2) (S5).

  Here, the log record deletion unit 112 first calculates the average difference MDF by the above equation (1).

  Here, n is the total number of sets, xi is the number of records in the set i, and xj is the number of records in the set j.

Next, the log record deletion unit 112 calculates the average value μ using the above equation (2).
Here, n is the total number of sets, and x1, x2,... Are the numbers of records in the first, second,.

Next, the log record deletion unit 112 calculates the Gini coefficient GC2 using Equation (4).
GC2 = MDF / (2 × μ) (4)
If the Gini coefficient GC2 is greater than the threshold value GCT2 (S5: YES), the log record deletion unit 112 searches the log record storage unit 110 for a set having a larger number of records than the predetermined number (hereinafter referred to as the threshold value LT2). For each set, records are deleted from the set so that the number of records in the set is equal to the threshold LT2 (S7).

  When the Gini coefficient GC2 is less than the threshold value GCT2 (S5: NO) or when the process of step S7 is finished, the log record deletion unit 112 records the service ID for each service ID in the log record storage unit 110. Find the average of the scores (average score).

  Next, the log record deletion unit 112 obtains the Gini coefficient GC3 indicating the imbalance of each average score, and determines whether the Gini coefficient GC3 is greater than a predetermined threshold (hereinafter referred to as threshold GCT3) (S9). ).

Here, the log record deletion unit 112 first calculates the average difference MDF by the above equation (1).
Here, n is the total number of average scores, xi is the average score i, and xj is the average score j.

Next, the log record deletion unit 112 calculates the average value μ using the above equation (2).
Here, n is the total number of average scores, and x1, x2,... Are the first, second,.

Next, the log record deletion unit 112 calculates the Gini coefficient GC3 using Expression (5).
GC3 = MDF / (2 × μ) (5)
If the Gini coefficient GC3 is greater than the threshold GCT3 (S9: YES), the log record deletion unit 112 determines the service ID corresponding to the average score that is greater than a predetermined point (hereinafter referred to as the threshold LT3) in the log record storage unit 110. A record including the above is retrieved, and the product of the average difference MDF calculated in step S9 and a predetermined coefficient is added to the score in the record (S11), and the process ends.

  If the Gini coefficient GC3 is less than the threshold value GCT (S9: NO), the log record deletion unit 112 ends the process.

  Returning to FIG. 1, the content recommendation unit 113 uses the log record storage unit 110 to obtain a content ID of content that the user would like to use within a predetermined maximum number or less, and includes the content ID. Is transmitted to a terminal used by the user.

  In addition, the content recommendation part 113 can use this operation | movement suitably, and can use the nonpatent literature 1, for example for calculating | requiring the content considered that a user likes at that time. For example, relationships between contents, users, and services are required, and contents are determined based on such relationships.

FIG. 9 is a flowchart showing the operation of the log deletion unit 114.
The log deletion unit 114 detects and deletes a log including a user ID and content ID that are no longer used from the log storage unit 101 (S21).

  Next, the log deletion unit 114 detects and deletes the record including the user ID deleted in step S21 from the integrated user ID database 108 (S23).

  Next, the log deletion unit 114 detects and deletes the record including the content ID deleted in step S21 from the integrated content ID database 109 (S25).

  Next, the log deletion unit 114 detects and deletes the record including the integrated user ID in the record deleted in step S21 or the integrated content ID in the record deleted in step S23 from the log record storage unit 110 (S27). ) Finish the process.

  Therefore, according to the present embodiment, in the recommendation server 1, the service ID, the content ID in the service, the content related information related to the content, the user ID in the service, the user related information related to the user, A user ID is read from each of a plurality of logs having the same user related information and different service IDs, and a user record including the user ID and the integrated user ID is integrated user The content ID is read from the user ID processing unit 106 stored in the ID database 108 and each of a plurality of logs having the same content related information and different service IDs, and the content record including the content ID and the integrated content ID is integrated. For each log of the content ID processing unit 107 and the log storage unit 101 to be stored in the content ID database 109, the integrated user ID database 108 is searched for a record including the user ID in the log and is read from the record. A log record for searching the record including the content ID in the log from the integrated content ID database 109 and storing the record including the integrated content ID read from the record and the service ID in the log in the log record storage unit 110 By configuring the log processing device including the generation unit 111, it is possible to provide a technique that eliminates inconvenience due to the difference between the user ID and the content ID in a plurality of services.

  For example, from the logs of services α and β used by a certain user A, the content of the service β can be recommended to the user B who has used the service β for the first time using the service α.

  The computer program for causing the computer to function as the log processing apparatus according to the present embodiment can be recorded on a computer-readable recording medium such as a semiconductor memory, a magnetic disk, an optical disk, a magneto-optical disk, or a magnetic tape. It can be widely distributed by being transmitted via a communication network such as the Internet.

DESCRIPTION OF SYMBOLS 1 ... Recommendation server 101 ... Log storage part 102, 103 ... Pre-processing part 104, 105 ... Service number storage part 106 ... User ID processing part 107 ... Content ID processing part 108 ... Integrated user ID database 109 ... Integrated content ID database 110 ... Log record storage unit 111 ... Log record generation unit 112 ... Log record deletion unit 113 ... Content recommendation unit 114 ... Log deletion unit

Claims (3)

A log processing device for processing a log accumulated to recommend content to a destination user when content can be transmitted in each of a plurality of services,
The service ID of the service indicating that any content of any service has been transmitted for any user, the content ID of the content at the service, content-related information related to the content, the service A log storage unit for storing a log including the user ID of the user and user-related information related to the user;
A log record storage unit for storing a record generated for each log of the log storage unit;
An integrated user ID database in which user records including a plurality of user IDs indicating the same user in a plurality of services are stored;
An integrated content ID database in which content records including a plurality of content IDs indicating the same content in a plurality of services are stored;
A user ID processing unit that reads a user ID from each of a plurality of logs having the same user-related information and different service IDs, and stores a user record including the user ID and the integrated user ID in the integrated user ID database;
A content ID processing unit that reads a content ID from each of a plurality of logs having the same content-related information and different service IDs, and stores a content record including the content ID and the integrated content ID in the integrated content ID database;
For each log in the log storage unit, a record including a user ID in the log is searched from the integrated user ID database, and an integrated user ID read from the record, and a content ID in the log from the integrated content ID database A log record generating unit that searches for a record to be included and generates a record including an integrated content ID read from the record and a service ID in the log, and stores the record in the log record storage unit Processing equipment. An operation method of a log processing apparatus for processing logs accumulated in order to recommend content to a destination user when content can be transmitted in each of a plurality of services,
The log processing device includes:
The service ID of the service indicating that any content of any service has been transmitted for any user, the content ID of the content at the service, content-related information related to the content, the service A log storage unit for storing a log including the user ID of the user and user-related information related to the user;
A log record storage unit for storing a record generated for each log of the log storage unit, and an integrated user ID database in which user records including a plurality of user IDs indicating the same user in a plurality of services are stored;
An integrated content ID database in which content records including a plurality of content IDs indicating the same content in a plurality of services are stored;
With
The operation method is as follows:
The user ID processing unit of the log processing device reads a user ID from each of a plurality of logs having the same user related information and different service IDs, and a user record including the user ID and the integrated user ID is stored in the integrated user ID. Memorize it in the database,
The content ID processing unit of the log processing device reads the content ID from each of a plurality of logs having the same content-related information and different service IDs, and the content record including the content ID and the integrated content ID is the integrated content ID. Memorize it in the database,
The log record generation unit of the log processing device searches for a record including the user ID in the log from the integrated user ID database for each log of the log storage unit, and reads the integrated user ID read from the record, the integrated Searching the record including the content ID in the log from the content ID database, generating a record including the integrated content ID read from the record and the service ID in the log, and storing the record in the log record storage unit. An operating method of a log processing apparatus.   A computer program for causing a computer to function as the log processing apparatus according to claim 1.

Images