JP5642228B2 - Correlation value calculation system, determination system, correlation value calculation method, determination method, and computer program - Google Patents

Description

  The present invention relates to a technique for acquiring a correlation between items related to an operation history based on a user's past operation history.

Conventionally, a recommendation technique for recommending (recommending) a store or a product highly relevant to the user using the purchase history of the user has been proposed. When selecting a store or product to recommend, for example, collaborative filtering is used.
Collaborative filtering is performed as follows. First, another person with a purchase history similar to the user to be recommended (hereinafter referred to as “recommended user”) is searched. Then, among the stores and products included in the searched purchase history of the others, the stores and products that have not been purchased by the recommending user are selected as stores and products to be recommended.
Conventionally, when making a recommendation using collaborative filtering, there is a problem that an appropriate recommendation cannot be made if the purchase history of the store or the product is small. For such a problem, a technique for recommending after integrating IDs in a plurality of services has been proposed (see Patent Document 1).

JP2013-12143A

However, collaborative filtering has a problem in that a product already known to a recommended user is recommended. There is also a problem that only similar products are always recommended to recommended users. For this reason, the user's willingness to purchase is not efficiently improved.
Such a problem is not a problem limited to the recommendation process of products using the user's purchase history, but is a common problem in general techniques for determining the correlation between items (for example, products) based on various operation histories of the user It is. That is, there is a demand for a technique for determining a correlation between items based on various operation histories of users by a new method that has not existed before.

  In view of the above circumstances, an object of the present invention is to provide a technique for acquiring correlation between items based on various operation histories of users.

One aspect of the present invention, based on the operation history indicating a history of user operations, a feature quantity acquisition unit that acquires a feature quantity for each item and is analyzed items were analyzed, before Symbol analysis target item A correlation value calculation unit that calculates a correlation value of the feature amount in a combination, and the feature amount acquisition unit is configured to analyze each analysis target based on the operation history of the user whose analysis target item appears in the operation history. a value representing the frequency of occurrence of the item is a correlation value calculation system that obtained as the feature amount.

  One aspect of the present invention is the above-described correlation value calculation system, wherein the feature amount acquisition unit acquires the feature amount based on the operation history of a user whose analysis target item appears in the operation history.

One aspect of the present invention is to select, as a first candidate item, the above-described correlation value calculation system, the analysis target item according to the transfer or borrowing of a processing target user, and another analysis target item having a high correlation value, Based on the transfer history that represents the past transfer or borrowing history of the user to be processed, the analysis target item that is less likely to be preferred or interested by the user to be processed is selected as a second candidate item. And a determination unit that selects the analysis target item included in duplicate in the first candidate item and the second candidate item.
One aspect of the present invention is the above-described determination system, in which the determination unit recommends the analysis target item included in the first candidate item and the second candidate item to the user to be processed. Select the item to be used.

One aspect of the present invention, based on the operation history indicating a history of user operations, a feature amount acquisition step of acquiring a feature quantity for each item and is analyzed items were analyzed, before Symbol analysis target item a correlation value calculation step of calculating a correlation value of the feature amount in the combination, were perforated in the feature amount obtaining step, based on the operation history of the user that the analysis target item appearing in the operation history, each analysis This is a correlation value calculation method for acquiring a value representing the appearance frequency of the target item as the feature amount .

In one aspect of the present invention, each step in the above correlation value calculation method, the analysis target item appearing in the operation history of the user to be processed, and another analysis target item having a high correlation value are used as first candidate items. An analysis target item that is selected and is less likely to be preferred or interested by the processing target user based on a transfer history that represents a past transfer or borrowing history of the processing target user is selected as a second candidate item. A determination step of selecting and selecting the analysis target item included in the first candidate item and the second candidate item as an item to be recommended to the processing target user.

  One aspect of the present invention is a computer program for causing a computer to function as the correlation value calculation system.

  One embodiment of the present invention is a computer program for causing a computer to function as the determination system.

  According to the present invention, it is possible to acquire correlation between items based on various operation histories of users by a new method.

It is a schematic block diagram which shows the function structure of the correlation value calculation system 10 which is 1st embodiment of this invention. It is a figure which shows the specific example of the table memorize | stored by the operation history memory | storage part. 4 is a flowchart showing an operation example of a correlation value calculation system 10. It is a figure which shows the specific example of the correlation value calculated by the correlation value calculation system. It is a schematic block diagram which shows the function structure of the recommendation determination system 20 which is 2nd embodiment of this invention. It is a figure which shows the specific example of the table memorize | stored by the transfer history storage part. 4 is a flowchart illustrating a specific example of a recommendation process of the recommendation determination system 20. It is a figure which shows the outline of the recommendation item selected by the recommendation determination part 209. FIG. It is a schematic block diagram which shows the function structure of the recommendation determination system 20a which is 3rd embodiment of this invention.

[First embodiment]
FIG. 1 is a schematic block diagram showing a functional configuration of a correlation value calculation system 10 according to the first embodiment of the present invention. The correlation value calculation system 10 is configured by one or a plurality of information processing apparatuses. For example, when the correlation value calculation system 10 is configured by a single information processing apparatus, the information processing apparatus includes a CPU (Central Processing Unit) connected via a bus, a memory, an auxiliary storage device, and the like, and a correlation value calculation program Execute. By executing the correlation value calculation program, the information processing apparatus functions as an apparatus including the operation history storage unit 101, the control unit 102, the feature amount acquisition unit 103, the feature amount storage unit 104, and the correlation value calculation unit 105. All or some of the functions of the correlation value calculation system 10 may be realized using hardware such as an ASIC (Application Specific Integrated Circuit), a PLD (Programmable Logic Device), or an FPGA (Field Programmable Gate Array). good. The correlation value calculation system 10 may be realized by dedicated hardware. The correlation value calculation program may be recorded on a computer-readable recording medium. The computer-readable recording medium is a portable medium such as a flexible disk, a magneto-optical disk, a ROM, a CD-ROM, a semiconductor storage device (for example, SSD: Solid State Drive), a hard disk built in a computer system, or a semiconductor storage. A storage device such as a device. The correlation value calculation program may be provided via a telecommunication line.

The operation history storage unit 101 is configured using a storage device such as a magnetic hard disk device or a semiconductor storage device. The operation history storage unit 101 stores a history of user's past operations for each user. The operation history stored in the operation history storage unit 101 includes a plurality of items. An item can be an analysis target or an element of a feature vector, and is, for example, a product name, a brand name, a store name, or a search keyword.
The control unit 102 controls the operation of the correlation value calculation system 10.

Based on the operation history stored in the operation history storage unit 101, the feature amount acquisition unit 103 obtains a feature amount for each item (hereinafter referred to as “analysis target item”) that is an analysis target regarding the correlation value. get. Which item is the analysis target item is appropriately set by an administrator, a user, or the like of the correlation value calculation system 10.
The feature amount storage unit 104 is configured using a storage device such as a magnetic hard disk device or a semiconductor storage device. The feature amount storage unit 104 stores the feature amount acquired by the feature amount acquisition unit 103 for each analysis target item.

  The correlation value calculation unit 105 calculates the correlation value of the feature amount for each combination of analysis target items. The feature value correlation value is a value representing the degree of similarity of feature values. The correlation value calculation method applied to the correlation value calculation unit 105 may be any method.

(About operation history)
A specific example of the operation history will be described. The operation history may be, for example, a history of operations performed by the user on the network (hereinafter referred to as “net history”). The operation history may be, for example, a history of points where the user has moved in real space (hereinafter referred to as “movement history”).

More specific examples of net history are listed below.
The net history may be a keyword history (hereinafter referred to as “search log”) used when a user performs a search on the network. In this case, each keyword is an item.
The net history may be a history of websites accessed by a user using a browser. In this case, the URL (Uniform Resource Locator) of the accessed website, the keyword set in the tag of the accessed website, and the like are items.

The net history may be a word included in a sentence posted by a user on an SNS (Social Networking Service), a blog, or the like. In this case, each word included in the posted sentence is an item.
The net history may be a history of a website registered by the user as a favorite of the browser. In this case, the URL of the registered website, the keyword set in the tag of the registered website, and the like are items.

More specific examples of the movement history are listed below.
The movement history may be a latitude and longitude history (hereinafter referred to as “coordinate history”) of each point where the user has moved in the real space. For the coordinate history, for example, latitude and longitude values acquired at a predetermined timing by a terminal device carried by the user using a satellite positioning system may be used. In this case, items such as the latitude and longitude values themselves, each facility (commercial facility, public facility, etc.) existing at the point indicated by the latitude and longitude are items.

  The movement history may be a history of facilities visited by the user in real space (hereinafter referred to as “facility history”). The facility history may be a facility history registered by the user using a function for registering the location of the user in the SNS (for example, a function such as check-in). In this case, each facility is an item.

  The operation history need not be limited to those described above, and may be any history as long as it is a history of the user's past operations. The operation history may be configured using a plurality of specific examples among the specific examples described above. This completes the description of the operation history.

  FIG. 2 is a diagram illustrating a specific example of a table stored by the operation history storage unit 101. In the example illustrated in FIG. 2, the operation history storage unit 101 stores the date and time when the operation was performed for each item indicating the user's operation. If the operation history shown in FIG. 2 is a search log, the value at the top of the table is that the user searched at 13:31:21 on March 2, 2013 using the keywords A1 and A2. Indicates. The value in the second row from the top of the table indicates that the user has searched at 14:11:10 on March 2, 2013 using the keyword a. The operation history storage unit 101 stores a table as shown in FIG. 2 for each user.

FIG. 3 is a flowchart illustrating an operation example of the correlation value calculation system 10. Hereinafter, a specific example of the processing flow of the correlation value calculation system 10 will be described with reference to FIG.
First, the control unit 102 determines one item to be processed (hereinafter referred to as “processing target item”) from the analysis target items (step S101). For example, the control unit 102 may determine the processing target item from the analysis target items in order according to a predetermined order. The control unit 102 outputs the determined processing target item to the feature amount acquisition unit 103.

  Next, the feature amount acquisition unit 103 selects a user whose processing target item appears in the operation history (step S102). In this processing, the feature amount acquisition unit 103 may be configured to select a user if the processing target item appears even once, or select a user of an operation history in which the processing target item appears a predetermined number of times or more. It may be configured to. The user selected by the feature amount acquisition unit 103 is referred to as “selected user” in the following description.

  Next, the feature amount acquisition unit 103 acquires the frequency of appearance in the operation history of the selected user for each item (step S103). Items whose frequency is acquired are items other than the processing target items. For example, when the processing target item is “A1” in FIG. 2, the feature amount acquisition unit 103 determines the item “A2”, the item “a”, the item “C”, the item “A3”, the item “d”, and the like. The frequency of appearance in the operation history of the selected user is acquired. The frequency of appearance may be, for example, the number of appearances in the operation history, or may be a value in which the number of appearances is normalized according to the number of all items appearing in the operation history. It may be a value multiplied by a predetermined coefficient.

  Next, the feature quantity acquisition unit 103 generates a feature quantity vector whose component is the frequency of each acquired item (step S104). Then, the feature amount acquisition unit 103 records the generated feature amount vector in the feature amount storage unit 104 in association with the processing target item (analysis target item) (step S105).

  Next, the control unit 102 determines whether or not the calculation of the feature amount vector has been completed for all the analysis target items (step S106). When the analysis target item for which the feature quantity vector has not been calculated remains (NO in step S106), the control unit 102 returns to step S101 and selects a new analysis target item from which the feature quantity vector has not been calculated. Determine the items to be processed. After that, the feature quantity acquisition unit 103 performs the processing of steps S102 to S105, thereby calculating a feature quantity vector of a new processing target item.

  On the other hand, when the calculation of the feature amount vector has been completed for all the analysis target items (step S106—YES), the correlation value calculation unit 105 is based on the feature amount vector recorded in the feature amount storage unit 104. A correlation value is calculated for each combination of analysis target items (step S107). The correlation value calculation unit 105 may calculate correlation values for only a predetermined combination of analysis target items, or may calculate correlation values for all combinations of analysis target items.

  In the correlation value calculation system 10 configured as described above, it is possible to acquire a correlation between items (between analysis target items) based on various operation histories of the user. Hereinafter, the effect produced by the correlation value calculation system 10 will be described in detail with reference to FIG.

  FIG. 4 is a diagram illustrating a specific example of the correlation value calculated by the correlation value calculation system 10. In the table shown in FIG. 4, items “A”, “B”, “C”, “D”, “E” and items “a”, “b”, “c”, “d”, “e” , The calculated correlation value is shown for each combination with “f”. In this case, items “A”, “B”, “C”, “D”, “E”, “a”, “b”, “c”, “d”, “e”, “f” are analyzed. It is an item.

  In the example shown in FIG. 4, the item “A” and the item “a” have a high correlation value. That is, the feature amount vector obtained based on the operation history of the user having the operation history regarding the item “A” is similar to the feature amount vector obtained based on the operation history of the user having the operation history regarding the item “a”. Yes. Assuming that the operation history is information that expresses the interests, interests and preferences of each user, the selected users of the item “A” and the item “a” whose feature vectors are similar to each other It can be estimated that these groups (user groups) are similar in interest, interest, and preference. Therefore, it can be determined that the item “A” and the item “a” are highly correlated in that the interests, interests, and preferences of the user group are similar. Such processing makes it possible to acquire the correlation between items on an unprecedented basis.

<Application example>
The correlation value obtained by the correlation value calculation system 10 may be applied to various fields.
For example, when one analysis target item is a fashion model and the other analysis target item is a clothing / accessory store, the correlation value calculation system 10 outputs the correlation between the fashion model and the clothing / accessory store. It becomes possible to do. The correlation value calculation system 10 may output a fashion model having a high correlation value as a candidate for a fashion model to be adopted for a clothing / accessory store. Instead of the clothing / accessory retailer, clothing / accessory brands, magazines, TV programs, products / services, and the like may be used as analysis target items.

  For example, the correlation value calculation system 10 can output the correlation between stores by setting both analysis target items as stores. The correlation value calculation system 10 may output another store having a high correlation value to the store as a store to be recommended to a user who has an operation history regarding the store. Instead of a store, a product or service may be used as an analysis target item.

  For example, when one analysis target item is a company in one industry and the other analysis target item is a company in another industry, the correlation value calculation system 10 may output a correlation between companies in different industries. It becomes possible. Correlation value calculation system 10 may output a company with a high correlation value as a candidate for a company with which a company in one industry collaborates or cooperates.

<Modification>
In the example shown in FIG. 1, the correlation value calculation system 10 is implemented as a single device, but the correlation value calculation system 10 may be implemented as a plurality of devices. For example, a device including the operation history storage unit 101 and a device including the control unit 102, the feature amount acquisition unit 103, the feature amount storage unit 104, and the correlation value calculation unit 105 may be implemented as different devices. In this case, each device may be connected to be able to transmit data via a network.

  When the correlation value calculation unit 105 outputs a combination of analysis target items having a high correlation value, the correlation value calculation unit 105 also outputs an item having a high frequency value in one or both of the feature amount vectors of the two analysis target items having a high correlation value. It may be configured as follows. By outputting such items, it is possible to estimate a potential item that indicates a user's interest or interest in a combination of analysis target items having a high correlation value.

[Second Embodiment]
FIG. 5 is a schematic block diagram showing a functional configuration of the recommendation determination system 20 according to the second embodiment of the present invention. The recommendation determination system 20 is a system configured by applying the configuration of the correlation value calculation system 10 of the first embodiment. The recommendation determination system 20 determines items to be recommended (recommended) to a user who has given or borrowed. Hereinafter, the details of the recommendation determination system 20 will be described.

  The recommendation determination system 20 is configured by one or a plurality of information processing apparatuses. For example, when the recommendation determination system 20 is configured by a single information processing apparatus, the information processing apparatus includes a CPU, a memory, an auxiliary storage device, and the like connected by a bus, and executes a recommendation determination program. By executing the recommendation determination program, the information processing apparatus performs an operation history storage unit 201, a control unit 202, a feature amount acquisition unit 203, a feature amount storage unit 204, a correlation value calculation unit 205, a correlation value storage unit 206, and a transfer history record storage. It functions as an apparatus including the unit 207, the relevance calculation unit 208, and the recommendation determination unit 209. All or some of the functions of the recommendation determination system 20 may be realized using hardware such as an ASIC, PLD, or FPGA. Further, the recommendation determination system 20 may be realized by dedicated hardware. The recommendation determination program may be recorded on a computer-readable recording medium. The recommendation determination program may be provided via a telecommunication line.

  The operation history storage unit 201 is configured using a storage device such as a magnetic hard disk device or a semiconductor storage device. The operation history storage unit 201 stores a history of user's past operations for each user. The operation history stored in the operation history storage unit 201 includes a plurality of items.

The control unit 202 controls the operation of the recommendation determination system 20.
The feature amount acquisition unit 203 acquires a feature amount for each analysis target item based on the operation history stored in the operation history storage unit 201.

The feature amount storage unit 204 is configured using a storage device such as a magnetic hard disk device or a semiconductor storage device. The feature amount storage unit 204 stores the feature amount acquired by the feature amount acquisition unit 203 for each analysis target item.
The correlation value calculation unit 205 calculates the correlation value of the feature amount for each combination of analysis target items by the same processing as the correlation value calculation unit 105.

The correlation value storage unit 206 is configured using a storage device such as a magnetic hard disk device or a semiconductor storage device. The correlation value storage unit 206 stores the correlation value calculated by the correlation value calculation unit 205 for each combination of analysis target items.
The transfer history storage unit 207 is configured using a storage device such as a magnetic hard disk device or a semiconductor storage device. The transfer / acquisition history storage unit 207 stores the user's past transfer / acquisition history for each user. The transfer / borrowing history is information indicating a transfer / borrowing history performed by the user in the past. The borrowing history includes items such as products and services that the user has previously transferred or borrowed (rented), stores that the user has transferred products or services, and brands of products and services that the user has transferred or borrowed (rented). Is information. The transfer or borrowing may be paid or free.

  The degree-of-relevance calculation unit 208 determines a recommended user (a user to be recommended) and each analysis target item (for example, product, service, store, brand) based on the transfer history stored in the transfer history storage unit 207. The relevance of is calculated. The degree of association is a value indicating the high possibility that each analysis target item is preferred (interested) by the recommended user. The relevance calculation unit 208 may calculate the relevance using, for example, collaborative filtering. Also, the relevance calculation unit 208 may calculate the relevance using the Pearson product moment correlation coefficient. The technique used by the relevance calculation unit 208 to calculate the relevance may be any technique.

  The recommendation determination unit 209 determines items to be recommended to recommended users (hereinafter referred to as “recommendation items”). Specifically, the recommendation determination unit 209 is an analysis target item that has a high correlation value with an analysis target item related to assignment or borrowing performed by (or performed by) the recommended user, and has a low degree of association with the recommendation user. Select an item as a recommendation item. The analysis target item related to assignment or borrowing performed (or performed) by the recommended user may be determined by the recommendation determination unit 209 itself, or may be instructed to the recommendation determination unit 209 by another device.

  FIG. 6 is a diagram illustrating a specific example of a table stored by the transfer history storage unit 207. In the example shown in FIG. 6, the transfer / borrowing history storage unit 207 stores the date and time when the transfer or borrowing contract was made for each combination of store, brand, and product. For example, in the specific example of FIG. 6, the value at the top of the table indicates that the user purchased the product a of the brand A1 at the store A on March 2, 2013 at 13:51:29. The value in the second row from the top of the table indicates that the user purchased the product X of the brand X at the store A at 03:52:58 on March 2, 2013. The transfer history storage unit 207 stores a table as shown in FIG. 6 for each user.

  FIG. 7 is a flowchart illustrating a specific example of the recommendation process of the recommendation determination system 20. Hereinafter, a specific example of the flow of the recommendation process of the recommendation determination system 20 will be described with reference to FIG. Note that the recommendation determination system 20 calculates a correlation value for each combination of analysis target items in advance before executing the recommendation process, and registers the correlation value in the correlation value storage unit 206. The process in which the recommendation determination system 20 calculates the correlation value for each combination of analysis target items is the same as the process shown in the flowchart of FIG.

  First, the recommendation determination unit 209 determines a recommended user (step S201). The recommendation determination unit 209 may determine, for example, a user designated by a user or an administrator of the recommendation determination system 20 as a recommended user. The recommendation determination unit 209 may determine a user designated from another information processing apparatus (for example, a Web server) as a recommended user. The recommended user may be determined in any way.

  Next, the recommendation determination unit 209 selects an analysis target item related to assignment or borrowing performed (or performed) by the recommended user (step S202).

Next, the recommendation determination unit 209 selects an analysis target item having a high correlation value with the analysis target item selected in step S202 as a first candidate item (step S203). Whether or not the correlation value is high may be determined based on whether or not the correlation value is higher than a predetermined threshold value, for example.
Next, the degree-of-association calculation unit 208 calculates the degree of association between the recommended user and each analysis target item (step S204). Next, the recommendation determination unit 209 selects, as the second candidate item, an analysis target item having a low degree of association with the recommended user (Step S205).

  Next, the recommendation determination part 209 selects the analysis object item contained in a 1st candidate item and a 2nd candidate item as a recommendation item (step S206). Then, the recommendation determination unit 209 outputs the selected recommendation item (step S207).

  FIG. 8 is a diagram showing an outline of recommendation items selected by the recommendation determination unit 209. In FIG. 8, the vertical axis represents a correlation value with an analysis target item related to assignment or borrowing performed (or performed) by a recommended user. The higher the position, the higher the correlation value. In FIG. 8, the horizontal axis represents the degree of association with the recommended user. The closer to the right, the higher the degree of relevance. The correlation value and the degree of association represent the coordinates of each analysis target item. That is, the position of each analysis target item in the graph of FIG. 8 is uniquely determined based on the correlation value and the degree of association.

  The region shown by the graph in FIG. 8 is classified into four partial regions 90-1 to 90-4 based on the level of the correlation value and the level of the degree of association. An analysis target item having a low correlation value and relevance is located in the partial region 90-1. In the partial area 90-2, an analysis target item having a high correlation value and a low degree of association is located. An analysis target item having a high correlation value and high degree of association is located in the partial region 90-3. In the partial area 90-4, an analysis target item having a low correlation value and a high degree of association is located. The recommendation determination unit 209 selects, as a recommendation item, an analysis target item located in the partial region 90-2 among these partial regions.

The correlation value and the degree of association that are the boundaries between the partial areas 90-1 to 90-4 may be determined in advance. The correlation value and the degree of association serving as the boundary between the partial regions 90-1 to 90-4 are determined by the recommendation determination unit 209 based on the calculated statistical value (for example, average value, median value, etc.) of each value. May be. The recommendation value determination unit 209 determines the correlation value and the degree of association as a boundary between the partial regions 90-1 to 90-4 on the basis of whether the value is significantly high or low based on a predetermined significance level according to a statistical distribution. May be.
The correlation value that is the boundary between the partial regions 90-1 and 90-2 and the correlation value that is the boundary between the partial regions 90-3 and 90-4 may be the same value or different values. good. The degree of association that is the boundary between the partial areas 90-1 and 90-4 and the degree of association that is the boundary between the partial areas 90-2 and 90-3 may be the same value or different values. good.

  In the recommendation determination system 20 configured as described above, it is possible to determine an analysis target item (recommendation item) to be recommended to a recommendation user. Hereinafter, the effect produced by the recommendation determination system 20 will be described in detail.

  In the recommendation determination system 20, the correlation value between the analysis target items is calculated by the same technique as the correlation value calculation system 10 of the first embodiment. Then, another analysis target item having a high correlation value with the analysis target item related to assignment or borrowing performed (or performed) by the recommended user is selected as the first candidate item. Therefore, there is a high possibility that the first candidate items include analysis target items that are in line with interests, interests, and preferences that potentially exist for the user. In addition, an analysis target item having a low degree of relevance to the recommended user is selected as the second candidate item. Therefore, the second candidate item includes an analysis target item that is estimated not to be in line with the user's interest, interest, and preference based on the transfer history of the recommended user. That is, the second candidate item includes an analysis target item that has a low correlation with a product or service that the user has actually assigned or borrowed according to his / her own interests, interests, and preferences. Then, analysis target items included in the first candidate item and the second candidate item are selected as recommendation items. Therefore, for each user, it is possible to select, as a recommendation item, an analysis target item that has potential interest but is not connected to an action such as assignment or borrowing. By providing such recommended items to the user, it becomes possible to promote sales more accurately.

  By such processing, it is possible to solve the problem that a product already known to a recommended user is recommended, and the problem that only similar products are always recommended to a recommended user.

<Modification>
In the flowchart of FIG. 7, the process of step S204 may be executed before the process of step S202 or step S203, or may be executed in parallel with the process of step S202 or step S203.
The item (recommendation item) selected by the recommendation determination unit 209 through the above process does not necessarily have to be selected for the purpose of being recommended to the recommendation user. That is, the recommendation determination unit 209 may be configured as a function of simply selecting an item by executing the above process regardless of whether or not the user is recommended. In this case, the item selected by the recommendation determination unit 209 may be used in any way by another device, the user of the recommendation determination system 20, or the like.

[Third embodiment]
FIG. 9 is a schematic block diagram showing a functional configuration of a recommendation determination system 20a according to the third embodiment of the present invention. The recommendation determination system 20a is a modified version of the recommendation determination system 20 of the second embodiment. The recommendation determination system 20a calculates the correlation value between stores based on the search history managed by the search server 30a, calculates the degree of association regarding the store based on the transfer history, which is managed by the sales server 40a, and recommend item Select a store. That is, in the third embodiment, a plurality of stores are set as analysis target items. Details of the recommendation determination system 20a will be described below.

  The recommendation determination system 20a is configured by one or a plurality of information processing apparatuses. For example, when the recommendation determination system 20a is configured by a single information processing device, the information processing device includes a CPU, a memory, an auxiliary storage device, and the like connected by a bus, and executes a recommendation determination program. By executing the recommendation determination program, the information processing apparatus causes the control unit 202a, the feature amount acquisition unit 203a, the feature amount storage unit 204, the correlation value calculation unit 205, the correlation value storage unit 206, the relevance calculation unit 208a, and the recommendation determination unit 209a. It functions as a device provided with. All or some of the functions of the recommendation determination system 20a may be realized using hardware such as an ASIC, PLD, or FPGA. Further, the recommendation determination system 20a may be realized by dedicated hardware. The recommendation determination program may be recorded on a computer-readable recording medium. The recommendation determination program may be provided via a telecommunication line.

The control unit 202a controls the operation of the recommendation determination system 20a.
The feature amount acquisition unit 203a acquires a feature amount for each store that is an analysis target item based on the search history stored in the search history storage unit 201a of the search server 30a.
The feature amount storage unit 204, the correlation value calculation unit 205, and the correlation value storage unit 206 have the same configurations as those in the second embodiment.

  The degree-of-association calculation unit 208a calculates the degree of association between the recommended user and each store that is the analysis target item based on the transfer-and-borrowing history stored in the transfer-and-borrowing history storage unit 207a of the sales server 40a. The degree of relevance is a value indicating the high possibility that each store is preferred (interested) by the recommended user. The relevance calculation unit 208a may calculate the relevance using, for example, collaborative filtering. Also, the relevance calculation unit 208 may calculate the relevance using the Pearson product moment correlation coefficient.

  The recommendation determination unit 209a determines a store recommended for a recommended user (hereinafter referred to as “recommended store”) from the stores that are analysis target items. Specifically, the recommendation determination unit 209a selects, as a recommended store, a store having a high correlation value with a store related to transfer or borrowing performed (or performed) by the recommended user and having a low degree of association with the recommended user. .

  The search server 30a is a device that provides a search service by a search engine to a plurality of users. The search server 30a searches a website related to the keyword based on one or more keywords input by the user, and provides a search result. The search server 30a records a keyword history input by the user in the search history storage unit 201a for each user.

  The sales server 40a receives the actual transfer result, and records the transfer result history (transfer history) in the transfer history storage unit 207a for each user. For example, the sales server 40a may be a server that acquires information aggregated by a POS (Point Of Sale) system provided in a plurality of actual stores. For example, the sales server 40a may be a server that acquires information on sales performed in a virtual store built on a network.

  The search server 30a and the recommendation determination system 20a are communicably connected via a network. The sales server 40a and the recommendation determination system 20a are communicably connected via a network.

  In the recommendation determination system 20a, the analysis target item is a store. In the recommendation determination system 20a, a search log is used as an operation history. Therefore, in the recommendation determination system 20a, the correlation value for each store is calculated using each keyword recorded as a search log as an item. And the store contained in a 1st candidate item and a 2nd candidate item is selected as a recommendation item. Therefore, for each user, it is possible to select a store that can be estimated that there is a potential interest based on the net search log but has not yet been actually transferred or borrowed as a recommendation item. By providing such recommended items to the user, it becomes possible to promote sales more accurately.

<Modification>
The analysis target items in the recommendation determination system 20a of the third embodiment need not be limited to stores. For example, a product, brand, magazine, restaurant, or the like may be set as the analysis target item. A plurality of types of items (for example, products and stores) may be set as analysis target items.

In the example shown in FIG. 9, the recommendation determination system 20a is implemented as a single device, but the recommendation determination system 20a may be implemented as a plurality of devices. For example, a device including a control unit 202a, a feature amount acquisition unit 203a, a feature amount storage unit 204, a correlation value calculation unit 205, and a correlation value storage unit 206, and a device including a relevance degree calculation unit 208a and a recommendation determination unit 209a. Each may be implemented as a different device. In this case, each device may be connected to be able to transmit data via a network.
The embodiment of the present invention has been described in detail with reference to the drawings. However, the specific configuration is not limited to this embodiment, and includes designs and the like that do not depart from the gist of the present invention.

  The present invention is applicable to estimating the connection between items based on the correlation value of analysis target items.

DESCRIPTION OF SYMBOLS 10 ... Correlation value calculation system, 101 ... Operation history memory | storage part, 102 ... Control part, 103 ... Feature-value acquisition part, 104 ... Feature-value memory | storage part, 105 ... Correlation-value calculation part, 20, 20a ... Recommendation determination system (determination system) ), 201 ... operation history storage unit, 202, 202a ... control unit, 203, 203a ... feature amount acquisition unit, 204 ... feature amount storage unit, 205 ... correlation value calculation unit, 206 ... correlation value storage unit, 207, 207a ... Transfer history storage unit, 208, 208a ... relevance calculation unit, 209, 209a ... recommendation determination unit (determination unit), 30a ... search server, 40a ... sales server, 201a ... search history storage unit

Claims (7)

A feature amount acquisition unit that acquires a feature amount for each analysis target item that is an analysis target item based on an operation history representing a user's operation history;
A correlation value calculation unit for calculating a correlation value of the feature amount in a combination of the previous SL analysis target item,
Equipped with a,
The feature quantity acquisition unit, based on the operation history of the user that the analysis target item appearing in the operation history, the correlation value calculation system that obtains a value that represents the frequency of occurrence of each analysis target item as the feature quantity. A correlation value calculation system according to claim 1 ;
The analysis target item corresponding to the transfer or borrowing of the processing target user and another analysis target item having a high correlation value are selected as first candidate items, and the past transfer or borrowing history of the processing target user is represented. Based on the transfer history, the analysis target item that is less likely to be preferred or interested by the user to be processed is selected as a second candidate item, and overlaps the first candidate item and the second candidate item. A determination unit that selects the analysis target item included
A determination system comprising: The determination system according to claim 2 , wherein the determination unit selects the analysis target item included in the first candidate item and the second candidate item as an item to be recommended to the processing target user. A feature amount acquisition step for acquiring a feature amount for each analysis target item that is an analysis target item based on an operation history representing a user's operation history;
A correlation value calculation step of calculating a correlation value of the feature amount in a combination of the previous SL analysis target item,
I have a,
In the feature amount acquisition step, a correlation value calculation method for acquiring, as the feature amount, a value representing an appearance frequency of each analysis target item based on the operation history of the user in which the analysis target item appears in the operation history . Each step in the correlation value calculation method according to claim 4 ,
The analysis target item appearing in the operation history of the processing target user and another analysis target item having a high correlation value are selected as first candidate items, and the past assignment or borrowing history of the processing target user is represented. The analysis target item that is less likely to be preferred or interested by the user to be processed based on the transfer history is selected as a second candidate item, and is duplicated in the first candidate item and the second candidate item. A determination step of selecting the analysis target item included;
Determination method. A computer program for causing a computer to function as the correlation value calculation system according to claim 1 . Computer program for causing a computer to function as a determination system according to claim 2 or 3.

Images