JP6079479B2 - Information processing apparatus, information processing method, and information processing program - Google Patents

Description

  The present invention relates to a technique for recommending items to a user.

In recent years, with the advancement of digital technology and network technology, items such as digital contents and articles are distributed and sold using a network. Along with this, there is an increasing need for a technology that selects and provides information on items that are likely to be of interest to the user from among a large number of items and information that matches the purpose.
For example, there is a recommendation process using collaborative filtering as one of the techniques for recommending items. In the recommendation process using collaborative filtering, a large amount of usage information such as browsing history and usage history is accumulated, and the usage information is used to recommend an item set that is likely to be of interest to the user.
In addition, there is a growing need for a technique for recommending items even when there is little usage information, such as when services are started.
For example, Patent Literature 1 discloses a technique for recommending even when usage information is small.

JP2004-341484

Patent Document 1 can present an item even in a state where the usage information is small by combining an item related to an item accessed by the user and an item related to a large number of items accessed by the user in the past.
However, in Patent Document 1 described above, since only the usage information is used, an item that has never been accessed by the user is not recommended. In other words, since it is difficult to recommend various items, the user may feel that “only similar items are recommended” and lose interest in the recommendation.

  An object of this invention is to provide the technique which can recommend various items.

In order to solve the above-described problems of the conventional technology, the present invention stores usage information in which a usage subject identifier for identifying a usage subject is associated with an item identifier for identifying an item used by the usage subject. An attribute identifier based on the usage information and the attribute correspondence information, an item information storage unit that stores attribute correspondence information in which a usage information storage unit, an item identifier, and an attribute identifier that identifies an item attribute are associated with each other. And an attribute recommendation value that is a recommended value between the user entity identifier and a recommended attribute that is an attribute identifier related to the user entity identifier is selected for each user entity identifier based on the calculated attribute recommendation value A recommended attribute selection unit that creates recommended attribute information that associates each user entity identifier with the selected recommended attribute, and the recommendation attribute selection unit generates A recommended attribute information storage unit for storing attribute information, and recommended attribute information including a user entity identifier of one user entity are acquired from the recommended attribute information storage unit, and any of the attribute identifiers included in the acquired recommended attribute information Attribute correspondence information including such information is obtained from the item information storage unit, a recommended candidate item set that is a set of item identifiers included in the acquired attribute correspondence information is formed, and each item identifier included in the recommended candidate item set Provided is an information processing apparatus comprising a recommended item selection unit that calculates a recommended value with the one user and selects an item recommended for the one user based on the calculated recommended value To do.
Further, in order to solve the above-described problems of the conventional technology, the present invention provides usage information in which a usage subject identifier for identifying a usage subject is associated with an item identifier for identifying an item used by the usage subject. A usage information storage step for storing in the usage information storage unit; an item information storage step for storing in the item information storage unit attribute correspondence information in which an item identifier and an attribute identifier for identifying an item attribute are associated; Based on the information and the attribute correspondence information, an attribute recommendation value, which is a recommended value between the attribute identifier and the user entity identifier, is calculated, and each user entity identifier is used for each user entity identifier based on the calculated attribute recommendation value. Select recommended attributes, which are attribute identifiers related to the identifier, and create recommended attribute information that associates each user entity identifier with the selected recommended attributes. A recommended attribute selection step, a recommended attribute information storage step for storing the recommended attribute information generated in the recommended attribute selection step in a recommended attribute information storage unit, and a recommendation attribute information including a user entity identifier of one user entity A set of item identifiers acquired from the attribute information storage unit and acquired from the item information storage unit attribute correspondence information including any of the attribute identifiers included in the acquired recommended attribute information. Forming a recommended candidate item set, calculating a recommended value between each item identifier included in the recommended candidate item set and the one user, and recommending to the one user based on the calculated recommended value There is provided an information processing method comprising a recommended item selection step of selecting an item to be performed.
In order to solve the above-described problems of the conventional technology, the present invention associates a computer with a user entity identifier for identifying a user entity and an item identifier for identifying an item used by the user entity. A usage information storage step for storing usage information in the usage information storage unit; an item information storage step for storing in the item information storage unit attribute correspondence information in which an item identifier and an attribute identifier for identifying an item attribute are associated; Based on the use information and the attribute correspondence information, an attribute recommendation value that is a recommended value between the attribute identifier and the use subject identifier is calculated, and for each use subject identifier based on the calculated attribute recommendation value, A recommended attribute that is an attribute identifier related to the user entity identifier is selected, and each user entity identifier is associated with the selected recommended attribute. A recommendation attribute selection step for creating recommendation attribute information, a recommendation attribute information storage step for storing the recommendation attribute information generated in the recommendation attribute selection step in a recommendation attribute information storage unit, and a recommendation including a use subject identifier of one use subject Attribute information is acquired from the recommended attribute information storage unit, attribute correspondence information including any of the attribute identifiers included in the acquired recommended attribute information is acquired from the item information storage unit, and included in the acquired attribute correspondence information A recommended candidate item set that is a set of item identifiers to be calculated, a recommendation value between each item identifier included in the recommended candidate item set and the one user is calculated, and based on the calculated recommended value, the one An information processing program that executes a recommended item selection step for selecting an item recommended for a user To provide.

  According to the present invention, various items can be recommended.

It is a figure which shows the structure of the whole system in the 1st Embodiment of this invention. It is a block diagram which shows the structure of the information processing server apparatus 1 in the 1st Embodiment of this invention. It is a figure which shows the storage state of the user information storage part 131 in the 1st Embodiment of this invention. It is a figure which shows the storage state of the item information storage part 132 in the 1st Embodiment of this invention. It is a figure which shows the storage state of the utilization information storage part 133 in the 1st Embodiment of this invention. It is a figure which shows the storage state of the recommendation attribute information storage part 134 in the 1st Embodiment of this invention. It is a figure which shows the storage state of the attribute preference information storage part 135 in the 1st Embodiment of this invention. It is a figure which shows the storage state of the attribute priority value information storage part 136 in the 1st Embodiment of this invention. It is a figure which shows the storage state of the recommendation item information storage part 137 in the 1st Embodiment of this invention. It is a flowchart which shows the procedure of the recommendation attribute selection process in the 1st Embodiment of this invention. It is a flowchart which shows the procedure of the attribute preference degree calculation process in the 1st Embodiment of this invention. It is a flowchart which shows the procedure of the attribute priority value calculation process in the 1st Embodiment of this invention. It is a flowchart which shows the procedure of the recommendation item selection process in the 1st Embodiment of this invention. It is a table | surface which shows the recommendation attribute with respect to the user u3 in the 1st Embodiment of this invention. It is a figure which shows the image of the item set of the recommendation candidate item with respect to the user u3 in the 1st Embodiment of this invention.

  Hereinafter, an information processing apparatus, an information processing method, and a program according to the present invention will be described with reference to the accompanying drawings. The item in the present invention may be digital content such as text, voice, music, video, web page, and various articles, or may be information on financial products, real estate, and people. That is, the item in the present invention may be tangible or intangible, whether it is paid or free.

<First Embodiment>
Hereinafter, a first embodiment of the present invention will be described in detail with reference to the drawings. FIG. 1 is a configuration diagram of the entire system according to the first embodiment of the present invention. In the system according to the present embodiment, the information processing server device 1 and one or more terminal devices 3 (3a to 3n) are connected via a network 2. In each embodiment, only the information processing server device 1 may function as an information processing device, or the information processing server device 1 may function as an information processing device in cooperation with the terminal device 3.
The network 2 is a network such as the Internet, for example, and mediates exchange of information between the information processing server device 1 and the terminal device 3.
The terminal device 3 is a general computer provided with a CPU, RAM, ROM, hard disk drive, network interface, or the like, or a portable information terminal such as a mobile phone, a smartphone, or a tablet PC, and the terminal control unit 31 and the terminal communication unit 32. And a terminal input unit 33 and a terminal display unit 34.
The terminal communication unit 32 is a unit for communicating with the information processing server device 1 via the network 2.

The terminal input unit 33 is an interface for the user to operate the terminal device 3 such as a mouse or a keyboard if the terminal device 3 is a PC (Personal Computer) or a button if the terminal device 3 is a mobile phone.
The terminal display unit 34 is an interface for displaying various information and visually showing it to the user, such as a display.
The terminal control unit 31 performs usage information transmission processing, recommended item acquisition request transmission processing, and recommended item detailed information display processing via the network 2.
The usage information transmission process is a process of transmitting usage information to the information processing server device 1 when a user (user) who uses the terminal device 3 performs an operation such as browsing or using some item. . The usage information is information related to browsing and usage of the user's items, and at least a user identifier for uniquely identifying the usage user, or a terminal identifier for uniquely identifying the terminal device 3, and a target for browsing and usage. It includes an item identifier that uniquely identifies the item that has become, and usage date information that indicates when it was viewed or used. In the present embodiment, a case where a user identifier is used will be described as an example, but the same applies to a case where a terminal identifier is used. The term user entity identifier is used as a general term for the meaning of the user identifier and terminal identifier combined. In addition, a user or a terminal device used by the user is appropriately referred to as a user. Further, the use time information may be information indicating only the date, or may be information indicated in units of milliseconds, and may be information that can be understood up to a necessary unit according to the form of service. Moreover, you may give other information, such as evaluation information regarding a user's item, to utilization information as needed.

The recommended item acquisition request transmission process is information when the user performs an operation to acquire recommended item detailed information or when the terminal device 3 performs an operation of automatically acquiring recommended item detailed information. This is a process of transmitting a recommended item acquisition request to the processing server device 1. The recommended item detailed information is information on one or more items recommended to the user. The recommended item acquisition request includes at least the user identifier of the user. The recommended item acquisition request may be associated with an acquisition limit number that limits the number of recommended items to be acquired.
The recommended item detailed information display process is a process of receiving the recommended item detailed information corresponding to the transmitted recommended item acquisition request from the information processing server device 1 and displaying the received recommended item detailed information on the terminal display unit 34. is there.

The information processing server device 1 is a device that exchanges various information with the terminal device 3 via the network 2. The information processing server device 1 can be implemented as software (program) processing using a computer including a CPU, RAM, ROM, hard disk drive (HDD), network interface, and the like. In addition, you may comprise the information processing server apparatus 1 using a some computer. For example, in order to distribute the load, the distributed processing may be performed using a plurality of computers that perform processing corresponding to each unit of the information processing server device 1. Alternatively, the distributed processing may be performed in such a manner that the processing of a part of the information processing server device 1 is performed by a computer and the processing of another part is performed by another computer.
FIG. 2 is a configuration diagram of the information processing server apparatus 1 in the present embodiment. The information processing server device 1 in this embodiment includes an information processing server control unit 11, an information processing server communication unit 12, and an information processing server storage unit 13.
The information processing server communication unit 12 is a unit for communicating with the terminal device 3 via the network 2.

The information processing server storage unit 13 stores various data using a storage device such as a memory or an HDD. The information processing server storage unit 13 includes a user information storage unit 131, an item information storage unit 132, a usage information storage unit 133, a recommended attribute information storage unit 134, an attribute preference information storage unit 135, and attribute priority value information storage. And a recommended item information storage unit 137.
The user information storage unit 131 stores a plurality of user information. FIG. 3 is a diagram illustrating a storage state of the user information storage unit 131. The user information associates a user identifier (user_id) that uniquely identifies a user of the terminal device 3 and user attribute information (user_info), and is stored in a table format as shown in FIG. The user attribute information is information such as the user's name, age, gender, address (region), hobby, membership time (year, month, date, date, etc.), e-mail address, telephone number, and the like. If information processing server device 1 can purchase a product, it may include credit card information for paying the product.
The item information storage unit 132 stores a plurality of item identification information, attribute type information, attribute value information, and attribute correspondence information. FIG. 4 (FIGS. 4A to 4D) is a diagram illustrating a storage state of the item information storage unit 132.
The item identification information is an item identifier (item_id) for uniquely identifying an item and an item name (item_name) associated with each other, and is stored in a table format as shown in FIG. Note that item_id is unique and cannot be registered repeatedly.

The attribute type information is an association between an attribute type identifier (type_id) for uniquely identifying an item attribute type and an item attribute type name (type_name) as shown in FIG. Store in table format. The item attribute type corresponds to the item attribute item, for example, “creator”, “genre”, “production year”, “price”, and “user condition suitable for use”. Note that type_id is unique and cannot be registered repeatedly.
The attribute value information is an attribute type identifier (type_id), an attribute value identifier (attr_id) for uniquely identifying an attribute value corresponding to the attribute type identifier, and an attribute value (attr_name), It is stored in a table format as shown in FIG. For example, when the item is related to “book” and type_id is an identifier corresponding to “genre”, the attribute values are “fiction”, “love”, “cooking”, and the like. Note that the combination of (type_id, attr_id) is unique, and duplicate registration is not possible.
The attribute correspondence information is information in which attributes corresponding to items are associated with each other, and is associated with an item identifier (item_id), an attribute type identifier (type_id), and an attribute value identifier (attr_id). Store in the table format as shown in d). Note that the combination of (item_id, type_id, attr_id) is unique, and duplicate registration is not possible.

By using the tables of FIGS. 4A to 4D, a plurality of attributes of an item can be managed for each attribute type. The item identification information corresponding to an item, the attribute type information specified from the attribute correspondence information corresponding to the item, and the attribute value information are combined as item attribute information. A combination of type_name and attr_name is used as an item attribute. A combination of type_id and attr_id is used as an attribute identifier. In this embodiment, the attribute identifier is treated as a combination of the attribute type identifier and the attribute value identifier. However, in the case of a service in which the item attribute type does not exist (the attribute type identifier does not exist), the attribute value Only identifiers may be treated as attribute identifiers. In this case, it is not necessary to store attribute type information.
The usage information storage unit 133 stores a plurality of usage information. FIG. 5 is a diagram illustrating a storage state of the usage information storage unit 133. As shown in FIG. 5, a user identifier (user_id), an item identifier (item_id), and usage time information (dl_date) constituting usage information are associated with each other and stored in a table format. When other information is also included in the usage information, the storage format may be changed in accordance with the usage information so that all the information included in the usage information can be stored.
The recommended attribute information storage unit 134 stores a plurality of recommended attribute information. FIG. 6 is a diagram illustrating a storage state of the recommended attribute information storage unit 134. As shown in FIG. 6, the user identifier (user_id), the attribute type identifier (type_id), the attribute value identifier (attr_id), and the attribute recommendation value (a_value) in which the degree of recommendation is quantified are included in the recommended attribute information. Are stored in a table format. Note that the combination of (user_id, type_id, attr_id) is unique, and duplicate registration is not possible.

The attribute preference information storage unit 135 stores a plurality of attribute preference information. FIG. 7 is a diagram illustrating a storage state of the attribute preference information storage unit 135. As shown in FIG. 7, a user identifier (user_id), an attribute type identifier (type_id), an attribute value identifier (attr_id), and an attribute preference degree (t_value) obtained by quantifying the degree of preference of an attribute. ) And store them in a table format. Note that the combination of (user_id, type_id, attr_id) is unique, and duplicate registration is not possible.
The attribute priority value information storage unit 136 stores a plurality of attribute priority value information. FIG. 8 is a diagram illustrating a storage state of the attribute priority value information storage unit 136. As shown in FIG. 8, an item identifier (item_id), an attribute type identifier (type_id), an attribute value identifier (attr_id), and an attribute priority value obtained by quantifying the attribute priority ( p_value) and store them in a table format. Note that the combination of (item_id, type_id, attr_id) is unique, and duplicate registration is not possible.

  The recommended item information storage unit 137 stores a plurality of recommended item information. FIG. 9 is a diagram illustrating a storage state of the recommended item information storage unit 137. As shown in FIG. 9, the user identifier (user_id) of the recommendation target user configuring the recommended item information, the item identifier (item_id) of the item recommended to the recommendation target user (item_id), and the recommended item for the recommendation target user A recommendation value (r_value) obtained by quantifying the degree of recommendation is associated and stored in a table format. The combination of (user_id, item_id) is unique, and duplicate registration is not possible. By specifying the user identifier of the recommendation target user, all the recommended items corresponding to the recommendation target user can be acquired.

The information processing server control unit 11 performs overall control on each unit configuring the information processing server device 1. The information processing server control unit 11 includes a usage information registration unit 111, a recommended attribute selection unit 112, an attribute preference level calculation unit 113, an attribute priority value calculation unit 114, a recommended item selection unit 115, and recommended item detailed information creation. Part 116.
The usage information registration unit 111 receives usage information from the terminal device 3 via the network 2 and stores the received usage information in the usage information storage unit 133.
The recommended attribute selection unit 112 performs recommended attribute selection processing at each predetermined timing. As the predetermined timing, a predetermined time interval (for example, every 24 hours) may be used, or the usage information may be received a certain number of times. Also, the time interval may vary from Monday to Friday every 3 hours, Saturday every 6 hours, and Sunday every 12 hours. Also, the time interval may be changed according to the season, such as shortening the time interval in summer and increasing the time interval in winter.

The procedure of the recommended attribute selection process will be described with reference to the flowchart of FIG.
First, the recommended attribute selection unit 112 extracts all user identifiers included in the usage information from the usage information storage unit 133 without duplication (step 101). At this time, the usage information for extracting the user identifier includes usage time information corresponding to a range that goes back for a predetermined period (for example, predetermined by the service provider such as 3 months) from the point of the recommended attribute selection process. By limiting, the recommended attribute information may be selected only for the user who has been used most recently.
Next, the recommended attribute selection unit 112 selects, for example, one of the user identifiers extracted in step S101 in the order of extraction (step S102). The user corresponding to the user identifier selected here becomes the recommendation target user.

Next, the recommended attribute selection unit 112 uses the usage information stored in the usage information storage unit 133 to select a similar user for the user identifier of the recommendation target user selected in Step S102 (Step S102). S103).
In order to select similar users, the degree of similarity between the recommendation target user and other users may be calculated, and up to a predetermined number of users may be selected as similar users in descending order of similarity. May be selected as similar users.
As a method of calculating the similarity, for example, a Jaccard coefficient can be used based on usage information. When the Jaccard coefficient is used, the item set that the recommendation target user ur has used is I (ur), the item set that the other user u has used is I (u), and both the user ur and the user u When the number of items used is | I (ur) Ｉ I (u) |, and the number of items that at least one of the user ur and user u has used is | I (ur) ∪ I (u) | The similarity sim (ur, u) can be calculated by the equation (1).

Moreover, it can calculate similarly even if it uses the item attribute set corresponding to the item which the user used instead of the item set which the user used.
Further, by calculating the item preference level, the cosine distance and the Pearson product moment correlation coefficient can also be used for similarity calculation.

The item preference level is a value indicating the degree of preference of the user for the item, and can be calculated using usage information (first usage information) including the user identifier of the user who calculates the item preference level. Specifically, for example, it can be calculated from the number of first usage information or the usage time information included in the first usage information using the elapsed time until the attribute recommended value calculation processing is performed. . Moreover, if the usage information includes information on the evaluation of the user's item and the amount of money the user has paid for the item, the usage information can also be used for calculation.
In the case of using the cosine distance, for example, an item set that has been used by the recommendation target user ur is defined as I (ur), and the item preference level for the item i (i∈I (ur)) of the user ur is represented by vi (ur, i) Let I (u) be an item set that other users u have used, and let vi (u, i ′) be an item preference for item i ′ (i′∈I (u)) of other users u. ), An item that the user ur and the user u have used together is I (ur, u), and the item preference level of the user ur for the item ic (ic∈I (ur, u)) is vi (ur, ic) ), Where the item preference level for the item ic of another user u is vi (u, ic), the similarity sim (ur, u) can be calculated by the equation (2).

  When using the Pearson product-moment correlation coefficient, for example, let I (ur, u) be an item that the user ur and the user u have used together, and the number of items belonging to I (ur, u) Let n be the item preference for user ur's item ic (ic ∈ I (ur, u)), vi (ur, ic), and the item preference for other user u's item ic is vi (u, ic). At this time, the similarity sim (ur, u) can be calculated by Equation (3).

Similarity calculation can also be performed by using an attribute preference level indicating a degree of preference for a user's item attribute instead of the item preference level. With respect to the attribute preference level calculation method, the attribute preference level calculation method described in the attribute preference level calculation process by the attribute preference level calculation unit 113 described later may be used.
Of course, you may calculate the similarity between users using methods other than the above.
Next, the recommended attribute selection unit 112 acquires, from the item information storage unit 132, the recommended attribute selection target among the attribute type identifiers included in the attribute type information (step S104). Note that a service in which the item attribute type is managed by an attribute type identifier or the like needs to set two or more attribute type identifiers to be recommended attribute selection targets in order to realize the effect of the present invention. The attribute type identifier to be recommended attribute selection target may be set in advance by the service provider.
Next, the recommended attribute selection unit 112 selects, for example, one of the attribute type identifiers acquired in step S104 in the order of acquisition (step S105).
Next, the recommended attribute selection unit 112 first extracts item identifiers from the usage information storage unit 133 from usage information including any of the user identifiers of similar users selected in step S103 without duplication. Then, from the item information storage unit 132, the attribute correspondence information having any of the extracted item identifiers is identified from the attribute correspondence information having the attribute type identifier selected in step S105, and the attributes included in the identified attribute correspondence information The identifier is extracted as a recommended candidate attribute identifier without duplication (step S106).
Next, the recommended attribute selection unit 112 calculates an attribute recommendation value for each recommendation candidate attribute identifier extracted in step S106 (step S107).

Hereinafter, three types of attribute recommendation value calculation methods will be described.
The first method of attribute recommendation value calculation is a method in which the number of similar users using items belonging to the attribute identifier is totaled for each recommendation candidate attribute identifier, and the attribute recommendation value is calculated using the number. The first method for calculating the attribute recommendation value treats the influence (weight) of each similar user with respect to the attribute recommendation value all equally (same).
The second method of calculating the attribute recommendation value is a method of calculating the attribute recommendation value for each recommendation candidate attribute identifier using the sum of similarities of similar users using items belonging to the attribute identifier. In the second method of calculating the attribute recommendation value, the influence (weight) of each user with respect to the attribute recommendation value is changed, and a similar user having a higher similarity with the recommendation target user has a stronger influence.

The third method for calculating the attribute recommendation value is to calculate the attribute preference for the recommendation candidate attribute identifier for each similar user and use the item belonging to the attribute identifier for each recommendation candidate attribute identifier. This is a method of calculating an attribute recommendation value using the sum of attribute preference levels. In the third method of attribute recommendation value calculation, the influence (weight) of each user with respect to the attribute recommendation value is changed, and a user having a higher attribute preference degree for the item attribute of the attribute recommendation value calculation target has a stronger influence. It will be. The attribute preference level may be calculated using an attribute preference level calculation process by the attribute preference level calculation unit 113 described later.
Further, the attribute recommendation value may be calculated by combining the second method and the third method of attribute recommendation value calculation. Of course, the attribute recommendation value may be calculated using other methods.

Next, the recommended attribute selection unit 112 selects a recommended attribute according to the attribute recommended value calculated in step S107 (step S108). In order to select a recommendation attribute, a recommendation candidate attribute identifier having an attribute recommendation value equal to or greater than a predetermined value (previously set by the service provider) may be used as the recommendation attribute. Further, the recommended candidate attribute identifiers selected up to a predetermined number (set in advance by the service provider) in descending order of the attribute recommendation values may be used as the recommended attributes.
Next, for each recommended attribute selected by the recommended attribute selection unit 112 in step S108, in step S107, the user identifier selected in step S102, the combination of the attribute type identifier and attribute value identifier that are recommended attributes, and The recommended attribute information associated with the recommended attribute value calculated corresponding to the recommended attribute is stored in the recommended attribute information storage unit 134 (step S109).
Next, the recommended attribute selection unit 112 determines whether or not all attribute type identifiers have been selected in step S105 (step S110). If all are selected, the process proceeds to step S111, and if there is an unselected one, the process proceeds to step S105.
In step S111, the recommended attribute selection unit 112 determines whether all user identifiers have been selected in step S102. If all are selected, the series of processing from step S101 to step S111 is terminated, and if there are unselected ones, the process proceeds to step S102.

According to this procedure, in a service that manages item attribute types, recommended attribute information corresponding to two or more attribute type identifiers is stored in the recommended attribute information storage unit 134 for each user identifier. It will be.
Note that as the number of attribute type identifiers for which recommended attribute information is created increases, the number of item candidates for selecting recommended items can be increased.
In addition, in a service that does not manage the item attribute type (no attribute type identifier), the processes of step S106, step S107, and step S112 can be omitted.
Also, the usage information used to calculate the similarity and the attribute recommendation value falls within a range that goes back for a predetermined period (for example, set in advance by the service provider such as 3 months) from the point of the recommended attribute selection process. It is possible to select the recommended attribute information by using only the latest usage information by limiting to the information including the usage time information. In addition, by obtaining up to a predetermined number (for example, set in advance by the service provider such as 10) in the order close to the point of the recommended attribute selection process, the calculation amount when selecting recommended attribute information is reduced. It may be.
The above is the description of the recommended attribute selection process.

The attribute preference level calculation unit 113 performs an attribute preference level calculation process at each predetermined timing. The procedure of attribute preference level calculation processing will be described with reference to the flowchart of FIG.
First, the attribute preference level calculation unit 113 extracts all user identifiers included in the usage information from the usage information storage unit 133 without duplication (step S201).
Next, the attribute preference level calculation unit 113 selects, for example, one of the user identifiers extracted in step S201 in the order of extraction (step S202).
Next, the attribute preference degree calculation unit 113 acquires all the usage information having the user identifier selected in step S202 from the usage information storage unit 133 (step S203). At this time, the usage information to be acquired is limited to information including usage time information corresponding to a range that goes back a predetermined period (for example, preset by the service provider such as 3 months) from the time of attribute preference calculation processing. Thus, the attribute preference degree may be calculated using only the latest usage information.

Next, the attribute preference degree calculation unit 113 acquires all the attribute correspondence information having any item identifier included in the usage information acquired in step S203 from the item information storage unit 132 (step S204). Note that when acquiring the attribute correspondence information, the attribute correspondence information may be limited to only one including one or more attribute type identifiers set in advance by the service provider.
Next, the attribute preference level calculation unit 113 selects, for example, one of the attribute correspondence information acquired in step S204 in the order of acquisition (step S205).
Next, the attribute preference level calculation unit 113 calculates an attribute preference level for the attribute correspondence information selected in step S205 (step S206).

Eight types of attribute preference calculation methods will be described below.
The first method for calculating the attribute preference degree is a method for uniformly calculating the same value (for example, “1”) for any item attribute corresponding to any attribute correspondence information. The first method for calculating the attribute preference degree treats any item attribute equally.

  The second method for calculating the attribute preference degree is a method of calculating the value so that the smaller the number of items belonging to the item attribute corresponding to the attribute correspondence information selected in step S205 is, the larger the value is. Specifically, the number of item sets I (tx, txy) belonging to the attribute identifier (tx, txy) of the item attribute (a pair of the attribute type identifier tx and the attribute value identifier txy) is expressed as | I (tx, txy) | In this case, the attribute preference degree tv (u, tx, txy) corresponding to the user u may be calculated by the equation (4) using the constant α1 (> 0).

  The second method of calculating the attribute preference degree is such that the finer item attribute such that the number of items belonging to the item attribute is smaller increases the weight, and conversely the granularity such that the number of items belonging to the item attribute is larger. The coarser the item attribute, the smaller the weight. Of course, any formula may be used as long as the formula monotonously decreases with respect to the number of items belonging to the item attribute.

  The third method for calculating the attribute preference degree is a method of calculating the value so that the larger the number of attribute value identifiers corresponding to the attribute type identifier included in the attribute correspondence information selected in step S205 is, the larger the value is. . Specifically, when the number of attribute value identifier sets A (tx) corresponding to the attribute type identifier tx of the item attribute is | A (tx) |, the constant β1 (≧ 1) is used to correspond to the user u. The attribute preference level tv (u, tx, txy) to be calculated may be calculated by equation (5).

  The third method for calculating the attribute preference degree is to increase the weight for an attribute type for which a fine-grained attribute value is set such that the number of attribute values corresponding to the attribute type is large. The weighting is reduced for attribute types for which coarse-grained attribute values are set such that the number of corresponding attribute values is small. Of course, any expression may be used as long as the expression monotonically increases with respect to the number of attribute values corresponding to the attribute type.

  The fourth method for calculating the attribute preference degree is a method for calculating the attribute preference degree so that the larger the use of the item belonging to the item attribute corresponding to the attribute correspondence information selected in step S205 is, the larger the value is. Specifically, the number of usage information sets L (tx, txy) related to the use of any item in the item set I (tx, txy) belonging to the attribute identifier (tx, txy) of the item attribute is expressed as | L ( tx, txy) | can be calculated by replacing | A (tx) | in Equation (5) with | L (tx, txy) |. The fourth method of calculating the attribute preference degree can increase the weight of the item attribute to which many popular items that are used many times by many users belong. Of course, any expression may be used as long as the expression monotonically increases with respect to the number of items belonging to the item attribute.

  The fifth method for calculating the attribute preference degree is a method of calculating the value so that the more users who use items belonging to the item attribute corresponding to the attribute correspondence information selected in step S205, the larger the value. Specifically, there is no duplication from the set L (tx, txy) of usage information related to the use of any item included in the item set I (tx, txy) belonging to the attribute identifier (tx, txy) of the item attribute. When the number of extracted user sets U (tx, txy) is | U (tx, txy) |, by replacing | A (tx) | in Equation (5) with | U (tx, txy) | It can be calculated. The fifth method of calculating the attribute preference degree can increase the weighting of popular item attributes that are used by many users. Of course, any expression may be used as long as the expression monotonically increases with respect to the number of users using items belonging to the item attribute.

  The sixth method for calculating the attribute preference degree is a method in which the number of items belonging to the item attribute corresponding to the attribute correspondence information selected in step S205 used by the user corresponding to the user identifier selected in step S202 is large. In this way, the value is calculated so as to increase. Specifically, any one of the usage information sets L (u) related to item usage by the user u included in the item set I (tx, txy) belonging to the attribute identifier (tx, txy) of the item attribute. When | L (u, tx, txy) | is the number of sets L (u, tx, txy) of usage information related to the use of items, | A (tx) | (U, tx, txy) | The sixth method of attribute preference calculation allows the weighting to be increased as the user of the attribute preference calculation target uses more items belonging to the item attribute of the attribute preference calculation target. Of course, any expression may be used as long as the expression monotonically increases with respect to the number of items used belonging to the item attribute of the user.

  The seventh method for calculating the attribute preference degree has a large number of types of items belonging to the item attribute corresponding to the attribute correspondence information selected in step S205 and used by the user corresponding to the user identifier selected in step S202. This is a method of calculating the value so that the value becomes larger. Specifically, an item set I (u, tx) in which items belonging to the attribute identifiers (tx, txy) of item attributes are extracted without duplication from a set L (u) of usage information related to the use of items by the user u. , Txy) can be calculated by replacing | A (tx) | in Equation (5) with | I (u, tx, txy) | when | I (u, tx, txy) | The seventh method of calculating the attribute preference degree can increase the weight as the user of the attribute preference calculation target uses more types of items belonging to the item attribute of the attribute preference calculation target. Of course, any formula may be used as long as it is a formula that monotonously increases with respect to the number of types of items belonging to a certain item attribute used by the user.

  The eighth method of attribute preference calculation is the item preference of items belonging to the item attribute corresponding to the attribute correspondence information selected in step S205, which is used by the user corresponding to the user identifier selected in step S202. This is a calculation method that monotonously increases in accordance with the sum. Specifically, an item set I (u, tx) in which items belonging to the attribute identifiers (tx, txy) of item attributes are extracted without duplication from a set L (u) of usage information related to the use of items by the user u. , Txy) and for each item included in the extracted item set, after calculating the item preference degree vi (u, i) for the item i (∈I (u, tx, txy)) of the user u, Using the constant β2 (≧ 1), the attribute preference degree tv (u, tx, txy) corresponding to the user u may be calculated by the equation (6).

Regarding the item preference level, the item preference level calculation method described in step S103 of the above-described recommended attribute selection process may be used. In the eighth method for calculating the attribute preference degree, the weight of an item attribute that includes many items preferred by the user who is the target of attribute preference calculation can be increased. Of course, any formula may be used as long as it is a formula that monotonously increases with respect to the total item preference.
Moreover, you may calculate combining the 2nd-8th method of attribute preference degree calculation. For example, the second and fourth methods may be combined and calculated as in Expression (7) or Expression (8). Other than that, they may be freely combined so as not to change the properties of the respective calculation methods. The above is the description of the attribute preference degree calculation method.

Next, the attribute preference calculation unit 113 selects the user identifier selected in step S202, the attribute type identifier and the attribute value identifier included in the attribute correspondence information selected in step S205, and the attribute preference calculated in step S206. The attribute preference information associated with the degree is stored in the attribute preference information storage unit 135 (step S207).
Next, the attribute preference degree calculation unit 113 determines whether or not all attribute correspondence information has been selected in step S205 (step S208). If all are selected, the process proceeds to step S209, and if there is any unselected, the process proceeds to step S205.
In step S209, the attribute preference level calculation unit 113 determines whether all user identifiers have been selected in step S202. If all are selected, the series of processing from step S201 to step S209 is terminated, and if there is an unselected one, the process proceeds to step S202.
The above is the description of the procedure of attribute preference calculation processing.

The attribute priority value calculation unit 114 performs an attribute priority value calculation process at every predetermined timing. The attribute priority value calculation process will be described with reference to the flowchart of FIG.
First, the attribute priority value calculation unit 114 extracts all the item identifiers included in the usage information from the usage information storage unit 133 without duplication (step S301).
Next, the attribute priority value calculation unit 114 selects, for example, one of the item identifiers extracted in step S301 in the order of extraction (step S302).
Next, the attribute priority value calculation unit 114 acquires all the usage information having the item identifier selected in step S302 from the usage information storage unit 133 (step S303). At this time, the usage information to be acquired is limited to information including usage time information corresponding to a range that goes back a predetermined period (for example, preset by the service provider such as 3 months) from the time of attribute priority value calculation processing. Thus, the attribute priority value may be calculated using only the latest usage information.

Next, the attribute priority value calculation unit 114 acquires all attribute correspondence information having the item identifier selected in step S302 from the item information storage unit 132 (step S304). Note that when acquiring the attribute correspondence information, the attribute correspondence information may be limited to only one including one or more attribute type identifiers set in advance by the service provider.
Next, the attribute priority value calculation unit 114 selects, for example, one of the attribute correspondence information acquired in step S304 in the order of acquisition (step S305).
Next, the attribute priority value calculation unit 114 calculates an attribute priority value for the attribute correspondence information selected in step S305 (step S306).

Eight types of attribute priority value calculation methods will be described below.
The first method for calculating the attribute priority value is a method for uniformly calculating the same value (for example, “1”) for any item attribute corresponding to any attribute correspondence information. The first method of attribute priority value calculation treats any item attribute equally.
The second method for calculating the attribute priority value is a method of calculating the value so that the smaller the number of items belonging to the item attribute corresponding to the attribute correspondence information selected in step S305 is, the larger the value is. This can be calculated using a method similar to the second method of attribute preference calculation.

The third method for calculating the attribute priority value is a method of calculating the attribute priority value so that the larger the number of attribute value identifiers corresponding to the attribute type identifier included in the attribute correspondence information selected in step S305, the larger the value. . This can be calculated using a method similar to the third method of attribute preference calculation.
The fourth method for calculating the attribute priority value is a method for calculating the attribute priority value so that the larger the use of the item belonging to the item attribute corresponding to the attribute correspondence information selected in step S305, the larger the value. This can be calculated using a method similar to the fourth method of attribute preference calculation.
The fifth method for calculating the attribute priority value is a method of calculating the value so that the more users who use the item belonging to the item attribute corresponding to the attribute correspondence information selected in step S305, the larger the value. This can be calculated using a method similar to the fifth method of attribute preference calculation.

  The sixth method for calculating the attribute priority value is a method of calculating the attribute priority value so that the larger the number of items used corresponding to the item identifier selected in step S302, the larger the value. Specifically, if the number of sets L (i) of usage information related to the use of item i is | L (i) |, the attribute priority value corresponding to item i using constant β3 (≧ 1) What is necessary is just to calculate pv (i, tx, txy) by Formula (9).

  The sixth method of attribute priority value calculation can increase the weighting of popular items that are used many times by many users. Of course, any formula may be used as long as the formula increases monotonously with the number of items used.

  The seventh method for calculating the attribute priority value is a method of calculating the attribute priority value so that the value increases as the number of users using the item corresponding to the item identifier selected in step S302 increases. Specifically, when | U (i) | is the number of user sets U (i) extracted without duplication from the set L (i) of usage information related to the use of item i, | L (i) | can be calculated by replacing | U (i) |. The seventh method of attribute priority value calculation can increase the weight of popular items that are used by many users. Of course, any expression may be used as long as the expression monotonically increases with respect to the number of users used.

  The eighth method of attribute priority value calculation is a method of calculating the value so that the higher the total item preference of the user who uses the item corresponding to the item identifier selected in step S302, the higher the value. is there. Specifically, the user set U (i) extracted without duplication is extracted from the use information set L (i) related to the use of the item i, and the user u (∈U (i) included in U (i) is extracted. ), The item preference degree vi (u, i) of the user u for the item i is calculated, and the attribute preference degree pv (i, tx, txy) corresponding to the item i is calculated using the constant β4 (≧ 1). ) May be calculated by equation (10).

Regarding the item preference level, the item preference level calculation method described in step S103 of the above-described recommended attribute selection process may be used. The eighth method of attribute priority value calculation can increase the weighting of items that are preferred by many users. Of course, any formula may be used as long as it is a formula that monotonously increases with respect to the total item preference.
Further, the calculation may be performed by combining the second to eighth methods of attribute priority value calculation. For example, the combination of the second and fourth methods may be calculated by using Expression (7) and Expression (8) that combine the second and fourth methods of attribute preference calculation. Other than that, they may be freely combined so as not to change the properties of the respective calculation methods.
This completes the description of the attribute priority value calculation method.

Next, the attribute priority value calculation unit 114 associates the item identifier, the attribute type identifier, the attribute value identifier, and the attribute priority value calculated in step S306, which are included in the attribute correspondence information selected in step S305. The information is stored in the attribute priority value information storage unit 136 (step S307).
Next, the attribute priority value calculation unit 114 determines whether or not all attribute correspondence information has been selected in step S305 (step S308). If all are selected, the process proceeds to step S309, and if there is an unselected one, the process proceeds to step S305.
In step S309, the attribute priority value calculation unit 114 determines whether all item identifiers have been selected in step S302. If all are selected, the series of processing from step S301 to step S309 is terminated, and if there is an unselected one, the process proceeds to step S302.
The above is the description of the procedure of attribute priority value calculation processing.

The recommended item selection unit 115 performs a recommended item selection process at each predetermined timing. The procedure of the recommended item selection process will be described with reference to the flowchart of FIG.
First, the recommended item selection unit 115 extracts all user identifiers included in the recommended attribute information from the recommended attribute information storage unit 134 without duplication (step S401).
Next, the recommended item selection unit 115 selects one of the user identifiers extracted in step S401 as a recommendation target user identifier, for example, in the order of extraction (step S402).
Next, the recommended item selection unit 115 acquires all recommended attribute information having a user identifier that matches the recommendation target user identifier selected in step S402 from the recommended attribute information storage unit 134 (step S403).
Next, the recommended item selection unit 115 acquires all attribute preference information having a user identifier that matches the recommendation target user identifier selected in step S402 from the attribute priority value information storage unit 136 (step S404).

Next, the recommended item selection unit 115 identifies attribute correspondence information including any attribute identifier included in the recommended attribute information acquired in step S403 from the item information storage unit 132, and is included in the identified attribute correspondence information. All item identifiers to be extracted are extracted without duplication (step S405). The set of item identifiers extracted here becomes a recommended candidate item set.
Next, the recommended item selection unit 115 selects one of the item identifiers extracted in step S405 as a recommended candidate item identifier in the order of extraction (step S406).
Next, the recommended item selection unit 115 acquires all attribute priority value information having an item identifier that matches the recommended candidate item identifier selected in step S406 from the attribute priority value information storage unit 136 (step S407).
Next, the recommended item selection unit 115 calculates a recommended value of the recommended candidate item identifier selected in step S406 for the recommendation target user identifier selected in step S402 (step S408).

Seven types of recommended value calculation methods will be described below. In the description of the calculation method, the set of recommended attribute information corresponding to the recommendation target user ur acquired in step S403 is ST (ur).
The first method of calculating the recommended value is a method of calculating, as the recommended value, the number of recommended attribute information including any of the attribute identifiers corresponding to the recommended candidate item identifier selected in step S406. The first method for calculating the recommended value requires a small amount of calculation.

  The second method of calculating the recommended value is a method of calculating using the recommended attribute value of the recommended attribute information including any of the attribute identifiers corresponding to the recommended candidate item identifier selected in step S406. Specifically, from ST (ur), the number of attribute identifier sets TA (ur, ir) extracted from those including one of the attribute identifiers corresponding to the recommended candidate item ir is represented by | TA ( ur, ir) |, and an attribute recommendation value av (ur, tx, txy) with the target user ur corresponding to an arbitrary attribute identifier (tx, txy) included in TA (ur, ir) is recommended The value rv (ur, ir) can be calculated by the equation (11) using the constant α2 (0 ≦ α2 ≦ 1).

  The second method of calculating the recommended value is that any item attribute corresponding to the recommended candidate item is included in a lot of recommended attribute information, and the recommended attribute value increases as the attribute recommended value included in the recommended attribute information increases. Tends to be high. Equation (11) is the sum of the recommended attribute values when α2 is “0”, and the average of the recommended attribute values when α2 is “1”. That is, in order to increase the influence of the number of recommended attribute information, α2 may be brought close to “0”. Of course, other expressions may be used as long as the number of recommended attribute information and the attribute recommendation value increase monotonously.

The third method for calculating the recommended value is a calculation method that uses the attribute recommended value as in the second method for calculating the recommended value and changes the weight of the corresponding attribute recommended value for each attribute type identifier. . Specifically, a weighting coefficient m1 (tx) corresponding to the attribute type identifier tx (∈T) is previously set for each attribute type identifier included in the set T of attribute type identifiers that are calculation targets of attribute recommendation value calculation. After setting, a set ST (ur, ir) of recommended attribute information including any attribute identifier among the attribute identifiers corresponding to the recommended candidate item ir is formed from ST (ur).
Next, a subset T (ur, ir) of T extracted without duplication from ST (ur, ir) is formed. Then, among the recommended attribute information included in ST (ur, ir), a set TA (ur, ir, t ′) of attribute identifiers extracted from those including the attribute type identifier t′x (∈T (ur, ir)). x) is | TA (ur, ir, t′x) |, and attribute recommendation values corresponding to arbitrary attribute identifiers (t′x, t′xy) of TA (ur, ir, t′x) When av (ur, t′x, t′xy) is used, the recommended value rv (ur, ir) can be calculated by the equation (12) using the constant α3 (0 ≦ α3 ≦ 1).

  Note that the weighting coefficient m1 (tx) corresponding to the attribute type identifier tx (∈T) is calculated according to the granularity of the attribute type in the same manner as the third method for calculating the attribute priority value, so that It is desirable to absorb the granularity, but of course, the service providing side may freely set it regardless of the granularity between attribute types. In the third method of calculating the recommended value, in addition to the nature of the second method of calculating the recommended value, the weight of the attribute type is changed to determine which attribute type has priority for the attribute type. The provider can set it freely. Other formulas may be used as long as the recommendation value of an item having an item attribute with a high attribute recommendation value is calculated while reflecting the weight for the attribute type.

  The fourth method of calculating the recommended value is a calculation method that uses the attribute recommended value as in the second method of calculating the recommended value and changes the weight of the corresponding attribute recommended value for each attribute identifier. Specifically, the weighting coefficient m2 (tx, txy) corresponding to the attribute identifier (tx, txy) is set for each attribute identifier set TA including the attribute type identifier that is the calculation target of attribute recommendation value calculation. Above, any subset TA (ur, ir) extracted from ST (ur) without duplication from the recommended attribute information including any attribute identifier among the attribute identifiers corresponding to the recommended candidate item ir When the recommended attribute value corresponding to the attribute identifier (t′x, t′xy) is av (ur, t′x, t′xy), the recommended value rv (ur, ir) is a constant α4 (0 ≦ α4). ≦ 1) can be calculated by the equation (13).

  The weighting coefficient m2 (tx, txy) corresponding to the attribute identifier (tx, txy) is calculated according to the granularity of the item attribute in the same manner as the second method for calculating the attribute priority value, so However, of course, the service providing side may freely set it regardless of the granularity between item attributes. In the fourth method of calculating the recommended value, in addition to the property of the second method of calculating the recommended value, by changing the weight for each item attribute, the attribute recommended value corresponding to which item attribute is given priority. The provider can set it freely. Other formulas may be used as long as the recommendation value of an item having an item attribute with a high attribute recommendation value is calculated while reflecting the weighting for the attribute identifier.

  The fifth method of calculating the recommended value is a method of calculating using the attribute preference information for the recommendation target user acquired in step S404. Specifically, a set TA (ur, ir) of attribute identifiers obtained by extracting only the attribute identifiers of the recommendation target item ir without duplication from the set TT (ur) of attribute preference information acquired for the recommendation target user ur. When the attribute preference level corresponding to the attribute identifier (tx, txy) included in tv is tv (ur, tx, txy), the recommended value rv (ur, ir) using the constant α5 (0 ≦ α5 ≦ 1) Can be calculated by the equation (14).

  In the fifth method of calculating the recommended value, it is easy to preferentially select an item having an item attribute preferred by the recommendation target user as a recommended item by calculating the recommended value using the attribute preference degree. In addition, the property of the method used for attribute preference calculation is also reflected. Of course, other expressions may be used as long as they monotonically increase with respect to the sum of attribute preference levels.

  The sixth method of calculating the recommended value is a method of calculating using the attribute priority value information for the recommended candidate item acquired in step S407. Specifically, a set of attribute identifiers TA (ir, ir, which are extracted without duplication only from the attribute identifiers used by the user ur in the past from the set of attribute priority value information TP (ir) acquired for the recommended candidate item ir. ur), where the attribute priority value corresponding to the attribute identifier (tx, txy) is pv (ir, tx, txy), using the constant α6 (0 ≦ α6 ≦ 1), Can be calculated.

  In the sixth method of calculating the recommended value, the recommended value is calculated using the attribute priority value, so that the item having the item attribute preferred by the recommendation target user is given priority as in the fifth method of calculating the recommended value. It becomes easy to select as recommended item. In addition, the nature of the method used for attribute priority value calculation is also reflected. Of course, other expressions may be used as long as they monotonically increase with respect to the sum of attribute priority values.

  The seventh method of calculating the recommendation value is a method of calculating using the attribute preference information for the recommendation target user acquired in step S404 and the attribute priority value information for the recommendation candidate item acquired in step S407. Specifically, a set of attribute identifiers TA ′ (ur) extracted without duplication from the set of attribute preference information TT (ur) acquired for the recommendation target user ur and the attributes acquired for the recommendation candidate item ir. Attribute identifiers (t′x, t) included in a common part TA ′ (ur) ∩TA ′ (ir) with a set of attribute identifiers TA ′ (ir) extracted without duplication from the set of priority value information TP (ir) When the attribute preference level corresponding to 'xy) is av (ur, t'x, t'xy) and the attribute priority value is pv (ir, t'x, t'xy), the recommended value rv (ur, ir) may be calculated as an inner product value as shown in equation (16).

Further, it may be calculated as a cosine distance or a Pearson product moment correlation coefficient. According to the seventh method of calculating the recommended value, the recommended user calculates the recommended value using the attribute preference level and the attribute priority value, so that the recommended target user can obtain the recommended value more than the fifth method and the sixth method. It becomes easier to preferentially select items having preferred item attributes as recommended items. The properties of the method used for attribute preference calculation and the properties of the method used for attribute priority value calculation are also reflected.
Alternatively, the second to seventh methods for calculating recommended values may be combined. For example, the third and fourth methods may be combined and calculated as in Expression (17).

  Also, a new recommended value new_rv (ur, ir) is obtained using the recommended value rv2 (ur, ir) calculated by the second method and the recommended value rv5 (ur, ir) calculated by the fifth method. The recommended value may be calculated as in Expression (18) or Expression (19). Other than that, they may be freely combined so as not to change the properties of the respective calculation methods. The above is the description of the recommended value calculation method.

Next, it is determined whether the recommended item selection part 115 selected all the recommendation candidate item identifiers in step S406 (step S409). If all are selected, the process proceeds to step S410, and if there is an unselected one, the process proceeds to step S406.
In step S410, the recommended item selection unit 115 selects a recommended item according to the recommended value calculated in step S408. In order to select a recommended item, a recommended candidate item identifier having a recommended value equal to or greater than a predetermined value (previously set by the service provider) may be used as the recommended item. The recommended candidate item identifiers selected up to a predetermined number (preset by the service provider) in descending order of the recommended values may be used as the item identifiers of the recommended items.
Next, for each recommended item selected by the recommended item selection unit 115 in step S410, the recommendation target user identifier selected in step S402, the item identifier of the recommended item, and the recommended value calculated in step S408. The associated recommended item information is stored in the recommended item information storage unit 137 (step S411).

Next, it is determined whether the recommended item selection part 115 selected all the recommendation object user identifiers in step S402 (step S412). If all are selected, the series of processing from step S401 to step S412 is terminated, and if there is an unselected one, the process proceeds to step S402.
A specific example of the actual effect of this procedure will be described with reference to FIGS. FIG. 14 shows recommended attributes for the user u3. As shown in FIG. 14, it can be seen that the recommended attributes of the user u3 are (ta, ta7), (ta, ta13), (tb, tb4), (tb, tb8). FIG. 15 is an image of an item set that is a target of recommended candidate items for the user u3. As shown in FIG. 15, there are item sets corresponding to the recommended attributes of the user u3, I (ta, ta7), I (ta, ta13), I (tb, tb4), I (tb, tb8), and diagonal lines Is the item set I (ta, ta7) ∪I (ta, ta13) ∪I (tb, tb4) ∪I (tb, tb8) of recommended candidate items. In other words, the recommended item can be selected from items belonging to any of I (ta, ta7), I (ta, ta13), I (tb, tb4), and I (tb, tb8). Even if the usage information is small, it is possible to recommend an item, and it is possible to realize a variety of item recommendations by using various item attributes.

In the above procedure, when the attribute preference level is not used for calculating the recommended value, the process of step S404 can be omitted. Furthermore, since it is not necessary to create attribute preference information, the attribute preference level calculation unit 113 and the attribute preference information storage unit 135 are not necessary. Further, when the attribute priority value is not used for the recommendation value calculation, the process of step S407 can be omitted. Furthermore, since it is not necessary to create attribute priority value information, the attribute priority value calculation unit 114 and the attribute priority value information storage unit 136 are not necessary.
The above is the description of the recommended item selection process.

When the recommended item detailed information creation unit 116 receives a recommended item acquisition request from the terminal device 3 through the information processing server communication unit 12, the recommended item detailed information creation unit 116 creates recommended item detailed information by performing a recommended item detailed information creation process. The created recommended item detailed information is transmitted to the terminal device 3 through the server communication unit 12.
The recommended item detailed information creation process is a process for creating recommended item detailed information. Specifically, first, all recommended item information having a user identifier included in the received recommended item acquisition request is acquired from the recommended item information storage unit 137. Then, item information corresponding to the item identifier included in the acquired recommended item information is acquired from the item information storage unit 132, and recommended item detailed information is created using the acquired item information. When the recommended item acquisition request includes an acquisition limit number, when the recommended item information is acquired from the recommended item information storage unit 137, the acquisition limit number is acquired in descending order of recommended values. Good.

  As described above, by creating recommended item information based on the recommended attribute information, it is possible to recommend various items even with a small amount of usage information. For this reason, the user's interest in recommendation can be maintained and the use of the service can be promoted.

  The above-described embodiments of the present invention are illustrative examples, and are not limited to the above-described embodiments, and various modifications can be made without departing from the scope of the present invention. For example, you may combine each embodiment, a modification, etc. In addition, even if the configuration of the information processing apparatus (information processing server apparatus 1) is separated and communicated with the separate configuration via a network or the like, the functions of the information processing apparatus can be realized. Good.

  The present invention also includes a program for causing a computer to realize the functions of the respective units. These programs may be read from a recording medium and loaded into a computer, or may be transmitted via a communication network and loaded into a computer.

DESCRIPTION OF SYMBOLS 1 Information processing server apparatus 2 Network 3 Terminal apparatus 31 Terminal control part 32 Terminal communication part 33 Terminal input part 34 Terminal display part 11 Information processing server control part 111 Usage information registration part 112 Recommended attribute selection part 113 Attribute preference degree calculation part 114 Attribute Priority value calculation unit 115 Recommended item selection unit 116 Recommended item detailed information creation unit 12 Information processing server communication unit 13 Information processing server storage unit 131 User information storage unit 132 Item information storage unit 133 Usage information storage unit 134 Recommended attribute information storage unit 135 Attribute preference information storage unit 136 Attribute priority value information storage unit 137 Recommended item information storage unit

Claims (5)

A usage information storage unit for storing usage information in which a usage subject identifier for identifying a usage subject is associated with an item identifier for identifying an item used by the usage subject;
An item information storage unit that stores attribute correspondence information that associates an item identifier with an attribute identifier that identifies an item attribute;
Based on the usage information and the attribute correspondence information, an attribute recommendation value, which is a recommended value between the attribute identifier and the usage subject identifier, is calculated, and for each usage subject identifier based on the calculated attribute recommendation value, A recommended attribute selection unit that selects a recommended attribute that is an attribute identifier related to the user entity identifier, and creates recommended attribute information in which each user entity identifier is associated with the selected recommended attribute;
A recommended attribute information storage unit for storing recommended attribute information generated by the recommended attribute selection unit;
The recommended attribute information including the user entity identifier of one user entity is acquired from the recommended attribute information storage unit, and the attribute correspondence information including any of the attribute identifiers included in the acquired recommended attribute information is the item information storage unit A recommended candidate item set that is a set of item identifiers included in the acquired attribute correspondence information, and a recommended value between each item identifier included in the recommended candidate item set and the one user A recommended item selection unit that calculates and recommends an item recommended to the one user based on the calculated recommended value ;
The recommended item selection unit identifies one item identifier included in a recommended candidate item set corresponding to the one user, and an attribute identifier corresponding to the identified one item identifier is assigned to the one user. Calculating a recommended value between the one user and the one item identifier using the number included in the corresponding recommended attribute information;
An information processing apparatus characterized by that. A usage information storage unit for storing usage information in which a usage subject identifier for identifying a usage subject is associated with an item identifier for identifying an item used by the usage subject;
An item information storage unit that stores attribute correspondence information that associates an item identifier with an attribute identifier that identifies an item attribute;
Based on the usage information and the attribute correspondence information, an attribute recommendation value, which is a recommended value between the attribute identifier and the usage subject identifier, is calculated, and for each usage subject identifier based on the calculated attribute recommendation value, A recommended attribute selection unit that selects a recommended attribute that is an attribute identifier related to the user entity identifier, and creates recommended attribute information in which each user entity identifier is associated with the selected recommended attribute;
A recommended attribute information storage unit for storing recommended attribute information generated by the recommended attribute selection unit;
The recommended attribute information including the user entity identifier of one user entity is acquired from the recommended attribute information storage unit, and the attribute correspondence information including any of the attribute identifiers included in the acquired recommended attribute information is the item information storage unit A recommended candidate item set that is a set of item identifiers included in the acquired attribute correspondence information, and a recommended value between each item identifier included in the recommended candidate item set and the one user A recommended item selection unit that calculates and recommends an item recommended to the one user based on the calculated recommended value ;
The recommended item selection unit identifies one item identifier included in the recommended candidate item set corresponding to the one use subject, and the higher the attribute recommendation value of the recommended attribute corresponding to the identified one item identifier is, Calculating a recommended value between the one user and the one item identifier so that the value is high;
An information processing apparatus characterized by that. The recommended item selection unit adds the value obtained by adding the attribute recommendation values of a plurality of recommended attributes corresponding to the specified one item identifier by the value based on the number of the added recommended attributes. As a recommended value between the user and the one item identifier ,
The information processing apparatus according to claim 2. On the computer,
And use principal identifier for identifying the use mainly the usage information storing step the use information associating the item identifier Ru is stored in the usage information storage unit for identifying the items that are utilized by the use mainly
The item identifier, and the item information storing step in which the attribute identifier identifying the attribute of the item, Ru is storing attribute corresponding information associated with the item information storage section,
Based on the usage information and the attribute correspondence information, an attribute recommendation value, which is a recommended value between the attribute identifier and the usage subject identifier, is calculated, and for each usage subject identifier based on the calculated attribute recommendation value, A recommended attribute selection step of selecting recommended attributes that are attribute identifiers related to the user principal identifier, and creating recommended attribute information in which each user principal identifier is associated with the selected recommended attribute;
And recommended attribute information storing step of Ru is stored recommendation attribute information generated in the recommended attribute selection step in recommended attribute information storage unit,
The recommended attribute information including the user entity identifier of one user entity is acquired from the recommended attribute information storage unit, and the attribute correspondence information including any of the attribute identifiers included in the acquired recommended attribute information is the item information storage unit A recommended candidate item set that is a set of item identifiers included in the acquired attribute correspondence information, and a recommended value between each item identifier included in the recommended candidate item set and the one user And a recommended item selection step of selecting an item recommended for the one user based on the calculated recommended value ,
In the recommended item selection step, one item identifier included in the recommended candidate item set corresponding to the one user entity is specified, and an attribute identifier corresponding to the one item identifier specified is the one user entity Calculating a recommended value between the one user and the one item identifier using the number included in the corresponding recommended attribute information;
An information processing program characterized by that. On the computer,
And use principal identifier for identifying the use mainly the usage information storing step the use information associating the item identifier Ru is stored in the usage information storage unit for identifying the items that are utilized by the use mainly
And item identifier, and the item information storing step in which the attribute identifier identifying the attribute of the item, Ru is storing attribute corresponding information associated with the item information storage section,
Based on the usage information and the attribute correspondence information, an attribute recommendation value, which is a recommended value between the attribute identifier and the usage subject identifier, is calculated, and for each usage subject identifier based on the calculated attribute recommendation value, A recommended attribute selection step of selecting recommended attributes that are attribute identifiers related to the user principal identifier, and creating recommended attribute information in which each user principal identifier is associated with the selected recommended attribute;
And recommended attribute information storing step of Ru is stored recommendation attribute information generated in the recommended attribute selection step in recommended attribute information storage unit,
The recommended attribute information including the user entity identifier of one user entity is acquired from the recommended attribute information storage unit, and the attribute correspondence information including any of the attribute identifiers included in the acquired recommended attribute information is the item information storage unit A recommended candidate item set that is a set of item identifiers included in the acquired attribute correspondence information, and a recommended value between each item identifier included in the recommended candidate item set and the one user And a recommended item selection step of selecting an item recommended for the one user based on the calculated recommended value ,
In the recommended item selection step, one item identifier included in the recommended candidate item set corresponding to the one user is identified, and the recommended attribute attribute value corresponding to the identified one item identifier is higher, Calculating a recommended value between the one user and the one item identifier so that the value is high;
An information processing program characterized by that.