JP7200069B2 - Information processing device, vector generation method and program - Google Patents

Description

The present invention relates to an information processing device, a vector generation method, and a program for generating a feature vector used to specify actions recommended to a user.

Conventionally, there is known a system that recommends a POI (Point of Interest), which is a place of interest to a user, based on the action history of the user. Patent Literature 1 discloses a technique of calculating a degree of similarity between users based on the user's action history, and recommending to the user POIs of other users whose similarity indicates that the degree of similarity is similar to that of the user to be recommended. ing.

JP 2011-003151 A

With the technology according to Patent Literature 1, if the regions where the users acted are not common, the similarity may be calculated to be low. For example, if two users acted in different regions, the similarity may be calculated to be low even if the behavior tendencies of these two users are similar. For a user in a depopulated area with few behavior histories, there are few other users in the same area who have similar behavior tendencies to the user, and it may not be possible to identify other users with a high degree of similarity. In this case, there arises a problem that the accuracy of recommending POIs to users in depopulated areas decreases. Therefore, it is required to be able to generate information that is used to specify POIs to be recommended to the user, while diluting the locality and corresponding to the user's behavioral tendencies.

Therefore, the present invention has been made in view of these points, and provides an information processing device, a vector generation method, and a program capable of generating feature vectors corresponding to user behavior tendencies while diluting regional characteristics. intended to

The information processing device according to the first aspect of the present invention includes, for each action performed by a user, action identification information for identifying the action, and time-series data indicating the order in which the user performs each action. an acquisition unit that acquires, a selection unit that selects one category from among a plurality of categories to which each of the plurality of behavior identification information contained in the time-series data belongs, and the category selected by the selection unit Conversion including the action identification information belonging to a certain selected category and category information indicating the category to which the action identification information belonging to the other category different from the selected category and included in the time-series data belongs a data generation unit that generates data; for each of the action identification information or the category information included in the conversion data generated by the data generation unit, the action indicated by the action identification information or the action belonging to the category information; and an action vector generation unit that generates an action vector, which is a feature vector indicating a relationship with another action different from the action.

The action identification information included in the time-series data is selected from first action identification information that cannot specify the action position of the user's action and second action identification information that can specify the action position. and the data generation unit generates the action identification information corresponding to the first action identification information among the action identification information belonging to the selected category and the action identification information belonging to the other category. and the category information indicating the category to which the action identification information corresponding to the second action identification information belongs among the action identification information belonging to the other category.

The acquisition unit acquires a plurality of pieces of time-series data corresponding to each of a plurality of users, and the information processing device acquires the time-series data for each user corresponding to the time-series data acquired by the acquisition unit. A user vector generation unit may further include a user vector generation unit that generates a user vector, which is a feature vector indicating a feature of behavior of the user, based on the plurality of behavior vectors generated based on the conversion data.

The information processing device recommends, to the user corresponding to the user vector, an action corresponding to the action vector relatively close to the user vector among the plurality of action vectors generated by the action vector generation unit. It may further have a recommendation unit.
The recommendation unit may recommend the action corresponding to another user different from the user corresponding to the user vector.

The time-series data is time information indicating a time at which the user performed an action indicated by the action identification information, and further includes the time information associated with the action identification information. An action indicated by the action identification information associated with the time information indicating a time included in a predetermined period in the time-series data may be recommended.

The time-series data includes action identification information that can identify the action position where the user performed the action, and the recommendation unit recommends an action indicated by the action identification information in which the action position is included in a predetermined range. may
The recommendation unit may recommend an action indicated by the action identification information that is not included in the time-series data corresponding to the user corresponding to the user vector.

A vector generation method according to a second aspect of the present invention is executed by a computer, includes action identification information for identifying the action for each action performed by a user, and determines the order in which the user performs each action. a step of acquiring time-series data indicating, a step of selecting one of a plurality of categories to which each of the plurality of action identification information contained in the time-series data belongs, and a step of selecting the selected category Conversion data including the action identification information belonging to a selected category and category information indicating the category to which the action identification information belonging to the other category different from the selected category and included in the time-series data belongs and for each of the action identification information or the category information included in the generated conversion data, the action indicated by the action identification information or the action belonging to the category information, and other actions different from the action and generating an action vector that is a feature vector that indicates the relevance of .

A program according to a third aspect of the present invention provides a computer with, for each action performed by a user, action identification information for identifying the action, and time-series data indicating the order in which the user performs each action. an acquisition unit that acquires the above, a selection unit that selects one category from among a plurality of categories to which each of the plurality of behavior identification information contained in the time series data belongs, and the category selected by the selection unit Conversion data including the action identification information belonging to a selected category and category information indicating the category to which the action identification information belonging to the other category different from the selected category and included in the time-series data belongs and for each of the action identification information or the category information included in the conversion data generated by the data generation unit, the action indicated by the action identification information or the action belonging to the category information, and the action It functions as an action vector generation unit that generates an action vector, which is a feature vector indicating relevance to another action different from the action vector.

ADVANTAGE OF THE INVENTION According to this invention, it is effective in the ability to generate|occur|produce the feature vector corresponding to a user's action tendency, diluting regional characteristics.

It is a figure for demonstrating the outline|summary of an information processing system. It is a figure which shows the structure of an information processing apparatus. It is a figure which shows an example of the various data which an information processing apparatus uses. 4 is a flow chart showing the flow of processing of the information processing apparatus;

[Overview of information processing system S]
FIG. 1 is a diagram for explaining an outline of an information processing system S. As shown in FIG. The information processing system S is a system that is used to recommend POIs, which are places that the user U may be interested in. The information processing system S has a user terminal 1 and an information processing device 2 .

The user terminal 1 is a terminal used by the user U, and is a mobile terminal such as a smart phone, for example. The user terminal 1 is installed with, for example, application software (hereinafter referred to as “collection application”) for collecting information on actions performed by the user U. FIG. Actions performed by the user U include, for example, the user U visiting a store, the user U accessing a web page using the web browser of the user terminal 1, and the collection It is an action of operating other application software (for example, a search application capable of searching for stores, a browsing application capable of browsing news, etc.) that is different from the application.

By operating the collection application, the user terminal 1 generates time-series data indicating the behavior history of the user U as information related to the behavior performed by the user U, and transmits the generated time-series data to the information processing device 2 . The time-series data is data that includes action identification information for identifying each action performed by the user U and indicates the order in which the user U performed each action. The action identification information is, for example, the name of the store visited by the user U, the name of the web page accessed by the user U using the web browser of the user terminal 1 (for example, the title), and other information that the user U operated on the user terminal 1. It is the name of the application software and the like. Note that the action identification information is not limited to the name described above, and may be, for example, a string of alphanumeric characters, a URL of a web page, or the like as long as the action can be identified.

When the user terminal 1 receives a beacon ID transmitted by a short-range wireless communication device provided in the store, the collection application may specify the store associated with the beacon ID, or the store provides the beacon ID. When the user terminal 1 is connected to Wi-Fi (registered trademark), the store associated with the management ID of the access point of the Wi-Fi may be identified. The user terminal 1 transmits time-series data to the information processing device 2 at predetermined intervals (for example, every day).

The information processing device 2 is a device, such as a server, that recommends POIs to the user U who uses the user terminal 1 in which the collection application is installed.

In the example shown in FIG. 1, first, the information processing device 2 receives a Obtain time-series data ((1) in FIG. 1). The information processing device 2 selects one category from a plurality of categories to which each of the plurality of action identification information included in the acquired time-series data belongs ((2) in FIG. 1).

A category is information obtained by abstracting action identification information, and includes, for example, restaurants, cafes, pubs, supermarkets, and convenience stores. Specifically, when the action identification information can specify the action position where the user U has acted, the category identifies the action position where the user U has acted. It is information that has been converted in such a way that it becomes impossible to The category selected by the information processing device 2 is the type of POI recommended to the user U. FIG.

The information processing device 2 generates conversion data based on the acquired time-series data for each user U corresponding to the acquired time-series data ((3) in FIG. 1). The conversion data contains action identification information belonging to the selected category, which is the selected category, and category information that belongs to another category different from the selected category and indicates the category to which the action identification information contained in the time series data belongs. data that contains

The information processing device 2 generates an action vector based on the conversion data for each action identification information or category information included in the generated conversion data ((4) in FIG. 1). The action vector is a feature vector indicating the relationship between the action indicated by the action identification information or the action belonging to the category information and other actions different from the action. In this way, the information processing device 2 generates an action vector based on conversion data including action identification information belonging to the selected category and category information abstracted from other action identification information not belonging to the selected category. , it is possible to generate a feature vector corresponding to the behavioral tendency of the user U while diluting the locality.

Subsequently, the information processing device 2 generates a user vector based on a plurality of behavior vectors generated based on the conversion data corresponding to the user U for each user U corresponding to the acquired time-series data (Fig. (5) of 1). A user vector is a feature vector that indicates features of the user U's behavior. In the example shown in FIG. 1, the information processing device 2 generates a user vector corresponding to user U1, a user vector corresponding to user U2, and a user vector corresponding to user U3.

Then, for each user U, the information processing apparatus 2 recommends, as a POI, an action corresponding to an action vector relatively close to the user vector among the plurality of action vectors to the user U corresponding to the user vector (see FIG. 1 ). of (6)).

By doing so, the information processing device 2 can identify the POI corresponding to the user's U behavior tendency. As a result, the information processing device 2 can recommend POIs of other users U who are similar to behavior tendencies of the user U to be recommended, even when the regions where the plurality of users U have acted are different. As a result, the information processing apparatus 2 can reduce variations in the accuracy of recommending POIs to the user U even when the amount of information in the action history of the user U varies from region to region.
The configuration of the information processing device 2 will be described below.

[Configuration of information processing device 2]
FIG. 2 is a diagram showing the configuration of the information processing device 2. As shown in FIG. The information processing device 2 has a communication section 21 , a storage section 22 and a control section 23 . The communication unit 21 is an interface for connecting to a network, and includes, for example, a LAN (Local Area Network) controller.

The storage unit 22 is a storage medium such as a ROM (Read Only Memory), a RAM (Random Access Memory), and a hard disk. The storage unit 22 stores programs executed by the control unit 23 . The storage unit 22 stores information about the user U who uses the user terminal 1 in which the collection application is installed, and time-series data. The information about the user U includes user identification information for identifying the user U, for example. Details of the time-series data will be described later.

The control unit 23 is, for example, a CPU (Central Processing Unit). The control unit 23 functions as an acquisition unit 231, a selection unit 232, a data generation unit 233, an action vector generation unit 234, a user vector generation unit 235, and a recommendation unit 236 by executing programs stored in the storage unit 22. .

The acquisition unit 231 acquires, via the communication unit 21, time-series data including action identification information for identifying the action performed by the user U and indicating the order in which the user U performs each action. Acquire from terminal 1. The time-series data further includes time information indicating the time at which the user U performed the action indicated by the action identification information as information indicating the order in which the user U performed each action.

The acquiring unit 231 acquires the time-series data from the user terminal 1, for example, at predetermined intervals (for example, every day). The acquisition unit 231 acquires user identification information corresponding to the user U who uses the user terminal 1 from the user terminal 1 together with the time series data, and stores the time series data in the storage unit 22 in association with the user identification information.

FIG. 3 is a diagram showing an example of various data used by the information processing device 2. As shown in FIG. FIG. 3A shows time-series data corresponding to each user U, and each time-series data includes action identification information and time information associated with the action identification information. there is Note that the order of each action identification information included in the time-series data may indicate the order in which the user U performed each action.

Here, the action identification information included in the time-series data corresponds to either the first action identification information or the second action identification information. The first action identification information is action identification information that cannot specify the action position where the user U performed the action, for example, the name (title) of the web page accessed by the user U using the web browser of the user terminal 1 This is the name of other application software that the user U has operated on the user terminal 1 . The second action identification information is action identification information that can identify the action position, and is, for example, the name of the store that the user U has visited.

Returning to FIG. 2, the selection unit 232 selects one category from a plurality of categories to which each of the plurality of action identification information included in the time-series data belongs. The selection unit 232 selects one category by, for example, accepting one category specified by the administrator of the information processing device 2 . The selection unit 232 may randomly select one category from a plurality of categories to which each of the plurality of action identification information included in the time-series data belongs.

The data generation unit 233 includes the action identification information belonging to the selected category, which is the category selected by the selection unit 232, and the category to which the action identification information belonging to another category different from the selected category and included in the time-series data belongs. Generate conversion data including category information indicating Specifically, the data generation unit 233 sequentially selects one action identification information from a plurality of action identification information included in the time-series data, and performs conversion processing on the selected action identification information. generates conversion data including action identification information and category information. Data generation unit 233, as a conversion process, if the selected action identification information belongs to the selected category, does not convert the action identification information, if the selected action identification information belongs to another category different from the selected category, the Action identification information is converted into category information indicating a category to which the action identification information belongs. The data generation unit 233 generates, for each user U, conversion data corresponding to the user U, for example.

In the example shown in FIG. 3B, among actions A, B, and C, which are action identification information included in the time-series data, action A belongs to category X, action B belongs to category Y, and action C belongs to category Z. and the selection unit 232 selects category X from among categories X, Y, and Z. In this case, the data generation unit 233 does not convert the action A belonging to the category X which is the selected category, and converts the actions B and C belonging to categories different from the category X to the categories Y and Z to which they belong respectively. Convert to category information indicating In this way, the data generation unit 233 generates conversion data including action identification information indicating action A, category information indicating category Y, and category information indicating category Z by executing conversion processing.

When converting the action identification information into category information, the data generation unit 233 converts using conversion information in which the action identification information and the category information are associated, for example. When the action identification information indicates the name of the store visited by the user U, and the store is a chain store or a brand store, the data generation unit 233 generates the chain name included in the store name indicated by the action identification information. Alternatively, a brand name (part of the store name other than the position-specifiable part) may be used as a category for conversion. For example, if the store name of the chain store "ABC" is "ABC XYZ station front store", the data generation unit 233 generates the chain store "ABC ” as the category.

Even if the action identification information belongs to another category, the data generation unit 233 may not convert the action identification information into category information when the action identification information corresponds to the first action identification information. Specifically, the data generation unit 233, among the action identification information belonging to the selected category and the action identification information belonging to other categories, the action identification information corresponding to the first action identification information and the action belonging to the other category You may generate|occur|produce the conversion data containing the category information which shows the category to which the action identification information applicable to 2nd action identification information belongs among identification information.

More specifically, among the action identification information belonging to other categories, the data generation unit 233 does not convert the action identification information corresponding to the first action identification information into category information, and does not convert the action identification information corresponding to the second action identification information. The converted data may be generated by converting the action identification information to category information indicating the category to which the action identification information belongs. By doing so, the information processing device 2 can abstract by limiting to action identification information that can identify the action position where the user U has acted.

For each action identification information or category information included in the conversion data generated by the data generation unit 233, the action vector generation unit 234 determines whether the action indicated by the action identification information or the action belonging to the category information and the action different from the action. generates an action vector, which is a feature vector that indicates the relevance to the action of Specifically, for each action identification information or category information included in the converted data generated by the data generation unit 233, the action vector generation unit 234 generates the action identification information or the category information and a plurality of and the relationship between actions of .

A relationship between a plurality of actions indicates the degree of relevance between the first action and the second actions performed before and after the first action. For example, if all the conversion data of each user U indicate that an action belonging to category Y (action B, etc.) and an action belonging to category Z (action C, etc.) were performed after action A, The action vector generation unit 234 generates, as the action vector corresponding to the action A, an action vector indicating that the action belonging to category Y and the action belonging to category Z are highly related.

The action vector generation unit 234 uses a machine learning model that is learned to output an action vector by inputting an input vector, which is a feature vector generated based on action identification information or category information, for example, Generate action vectors. A machine learning model is a model learned using teacher data generated based on converted data.

The teacher data shown in FIG. 3(c) is data in which input information and prediction information are associated with each other. The input information is an input vector generated based on the action identification information indicating the first action performed by the user U or the category information indicating the category to which the first action belongs in the converted data generated by the data generation unit 233. The prediction information is a prediction vector that is a feature vector generated based on action identification information indicating a second action performed by the user U before or after the first action or category information indicating a category to which the second action belongs.

Specifically, first, the action vector generation unit 234 generates a machine learning model used to generate feature vectors. The action vector generator 234 generates teacher data based on the converted data generated by the data generator 233 . The action vector generation unit 234 causes the generated machine learning model to learn using the generated teacher data. Then, the action vector generation unit 234 inputs an input vector corresponding to the action identification information or the category information to the machine learning model for each action identification information or category information included in the conversion data, and the machine learning model outputs By acquiring the predicted vector, an action vector corresponding to the action identification information or the category information is generated.

For example, when the information processing device 2 executes processing for recommending a POI, the action vector generation unit 234 executes a series of processing from processing for generating a machine learning model to processing for generating an action vector. The action vector generation unit 234 stores in the storage unit 22 in advance a plurality of machine learning models that are machine learning models corresponding to each category that can be selected by the selection unit 232 and are learned using teacher data, When the selection unit 232 selects a category, a process of generating an action vector using a machine learning model corresponding to the selected category may be executed.

The user vector generation unit 235 generates a user vector, which is a feature vector that indicates features of behavior of the user U. FIG. Specifically, first, the acquisition unit 231 acquires a plurality of pieces of time-series data corresponding to each of the plurality of users U via the communication unit 21 . Then, for each user U corresponding to the time-series data acquired by the acquisition unit 231, the user vector generation unit 235 generates the user vector to generate The user vector generation unit 235, for example, calculates the average of a plurality of action vectors generated based on the conversion data corresponding to the user U, and generates the calculated result as the user vector.

The recommendation unit 236 recommends, as a POI, an action corresponding to an action vector relatively close to the user vector among the plurality of action vectors generated by the action vector generation unit 234 to the user U corresponding to the user vector. The action corresponding to the action vector is the action identified by the action identification information corresponding to the action vector, that is, the name of the store visited by the user U indicated by the action identification information. Specifically, the recommendation unit 236 selects a behavior relatively close to the user vector corresponding to the user U to be recommended from among the plurality of behavior vectors generated based on each of the plurality of behavior identification information belonging to the selected category. A vector is specified, and an action corresponding to the specified action vector is recommended to the user U to be recommended.

More specifically, for each action vector generated by the action vector generation unit 234 based on the action identification information belonging to the selected category, the recommendation unit 236 first determines the similarity between the action vector and the user vector. Calculate degrees. For example, the recommendation unit 236 calculates a higher degree of similarity as the behavior vector and the user vector are closer to each other. The recommendation unit 236 identifies an action vector corresponding to a relatively high degree of similarity from among the plurality of degrees of similarity calculated. The recommendation unit 236 identifies, for example, a plurality of action vectors included in a predetermined rank from the highest degree of closeness indicated by the plurality of similarities. The recommendation unit 236 may specify an action vector corresponding to a degree of similarity indicating that the degree of similarity is closer than a predetermined degree of similarity threshold among the plurality of degrees of similarity.

Then, the recommendation unit 236 recommends to the user U an action corresponding to the identified action vector. For example, the recommendation unit 236 notifies the user terminal 1 of the user U of the information indicating the POI via the communication unit 21, and displays the POI on the screen of the collection application installed on the user terminal 1. Recommend POIs to

The recommendation unit 236 may recommend an action corresponding to another user U different from the user U corresponding to the user vector. Specifically, the recommendation unit 236 selects, among the plurality of action vectors generated by the action vector generation unit 234, the action vector corresponding to the other user U and the action vector relatively close to the user vector. Actions are recommended to the user U corresponding to the user vector. The action vector corresponding to the other user U is the action vector generated based on the conversion data corresponding to the other user U. By doing so, the information processing device 2 can avoid recommending the POI of the user U to the user U to be recommended.

The recommendation unit 236 may recommend an action indicated by action identification information associated with time information indicating a time included in a predetermined period in the time-series data. The predetermined period may be a period specified by an administrator, or may be a period set in advance (for example, a period from the present to one month ago).

Specifically, the recommendation unit 236, among the plurality of action vectors generated by the action vector generation unit 234, based on the action identification information associated with the time information indicating the time included in a predetermined period in the time-series data An action corresponding to a generated action vector that is relatively close to the user vector may be recommended to the user U corresponding to the user vector. By doing so, the information processing device 2 can recommend to the user U an action performed within the specified period.

The recommendation unit 236 may recommend an action indicated by action identification information whose action position is included in a predetermined range. The predetermined range is, for example, prefectures or municipalities. The predetermined range may be a range specified by an administrator, or a user U specified by the action identification information corresponding to the second action identification information included in the time-series data corresponding to the user U to be recommended. may be within the range of action of

Specifically, the recommendation unit 236, among the plurality of action vectors generated by the action vector generation unit 234, based on the action identification information corresponding to the second action identification information in which the specified action position is included in a predetermined range An action vector generated by the method and corresponding to the action vector relatively close to the user vector may be recommended to the user U corresponding to the user vector. By doing so, the information processing device 2 can recommend to the user U actions performed within a predetermined range.

The recommendation unit 236 may recommend an action indicated by action identification information that is not included in the time-series data corresponding to the user U corresponding to the user vector. Specifically, the recommendation unit 236 selects other action identification information other than the action identification information included in the time-series data corresponding to the user U corresponding to the user vector among the plurality of action vectors generated by the action vector generation unit 234. An action vector generated based on the information and corresponding to an action vector relatively close to the user vector may be recommended to the user U corresponding to the user vector. By doing so, the information processing apparatus 2 can recommend an action that the user U to be recommended has never performed.

[Processing of information processing device 2]
Next, the processing flow of the information processing device 2 will be described. FIG. 4 is a flow chart showing the processing flow of the information processing device 2. As shown in FIG. In this flowchart, the administrator selects one category among a plurality of categories to which each of the plurality of behavior identification information contained in the plurality of time-series data acquired from each of the plurality of user terminals 1 by the acquisition unit 231 belongs. , is started when the type of POI to be recommended to the user U is specified.

The selection unit 232 selects one category specified by the administrator (S1). The data generation unit 233, for each user U corresponding to the time-series data acquired by the acquisition unit 231, the action identification information belonging to the selected category that is the category selected by the selection unit 232, and other categories different from the selected category Converted data including category information indicating the category to which the action identification information belonging and included in the time-series data belongs is generated (S2). The action vector generation unit 234 generates an action vector for each action identification information or category information included in the conversion data generated by the data generation unit 233 (S3).

The user vector generation unit 235 generates a user vector for each user U based on a plurality of action vectors generated based on the conversion data corresponding to the user U (S4). The recommendation unit 236 recommends, as a POI, an action corresponding to an action vector relatively close to the user vector among the plurality of action vectors generated by the action vector generation unit 234 to the user U corresponding to the user vector (S5). .

[Effects of this embodiment]
As described above, the information processing device 2 includes the action identification information belonging to the selected category, which is a category selected from among the plurality of categories to which each of the plurality of action identification information included in the acquired time-series data belongs, and the selection Generate conversion data that belongs to another category different from the category and contains category information that indicates the category to which the action identification information contained in the time series data belongs, and the action identification information or category included in the generated conversion data A behavior vector is generated for each piece of information based on the conversion data. In this way, the information processing device 2 generates an action vector based on conversion data including action identification information belonging to the selected category and category information abstracted from other action identification information not belonging to the selected category. , it is possible to generate a feature vector corresponding to the behavioral tendency of the user U while diluting the locality.

After generating the action vector, the information processing device 2 generates a user vector based on a plurality of action vectors generated based on the conversion data corresponding to the user U, and selects the user vector from among the plurality of action vectors. An action corresponding to a relatively close action vector is recommended to the user U corresponding to the user vector as a POI. By doing so, the information processing apparatus 2 recommends POIs of other users U that are similar to the behavioral tendencies of the user U to be recommended, even when the regions where the plurality of users U have acted are different. be able to. As a result, the information processing apparatus 2 can reduce variations in the accuracy of recommending POIs to the user U even when the amount of information in the action history of the user U varies from region to region.

[Modification]
In the above description, the information processing apparatus 2 selects the category specified by the administrator, but the present invention is not limited to this. For example, the information processing device 2 may select a category specified by the user U on the screen of the collection application installed in the user terminal 1 .

In this case, the information processing device 2 stores in advance in the storage unit 22 a plurality of machine learning models corresponding to each category that can be specified by the user U, which are learned using teacher data. , when the user U designates a category, a process of generating an action vector using a machine learning model corresponding to the designated category and a process of recommending a POI to the user U may be executed. If the information processing apparatus 2 has a high CPU performance and can perform processing of learning a machine learning model using teacher data at high speed, the processing of generating a machine learning model when the user U designates a category is performed. to the process of recommending the POI to the user U may be executed.

Although the present invention has been described above using the embodiments, the technical scope of the present invention is not limited to the scope described in the above embodiments, and various modifications and changes are possible within the scope of the gist thereof. be. For example, all or part of the device can be functionally or physically distributed and integrated in arbitrary units. In addition, new embodiments resulting from arbitrary combinations of multiple embodiments are also included in the embodiments of the present invention. The effect of the new embodiment caused by the combination has the effect of the original embodiment.

1 user terminal 2 information processing device 21 communication unit 22 storage unit 23 control unit 231 acquisition unit 232 selection unit 233 data generation unit 234 action vector generation unit 235 user vector generation unit 236 recommendation unit S information processing system U user

Claims (10)

an acquisition unit that acquires, for each action performed by a user, time-series data that includes action identification information for identifying the action and indicates the order in which the user performs each action;
From among a plurality of categories in which it is impossible to specify an action position where the user has performed an action, and to which each of the plurality of action identification information included in the time-series data belongs, the user A selection unit that selects one category to be recommended to the user as a type of POI (Point of Interest) that is a place that can be of interest;
The action identification information belonging to the selected category which is the category selected by the selection unit among the plurality of action identification information included in the time-series data is not converted, and the selected category among the plurality of action identification information A data generation unit that generates converted data obtained by converting the action identification information belonging to another category different from the category information to the category information indicating the other category;
Based on the order of each action performed by the user indicated by the time-series data, the action identification information is indicated for each of the action identification information or the category information included in the conversion data generated by the data generation unit. an action vector generation unit that generates an action vector that is a feature vector that indicates the relationship in order between an action or an action belonging to the category information and another action that is different from the action;
Information processing device having The action identification information included in the time-series data is selected from first action identification information that cannot specify the action position of the user's action and second action identification information that can specify the action position. falls under any of
The data generation unit includes, among the action identification information belonging to the selected category and the action identification information belonging to the other category, the action identification information corresponding to the first action identification information and the other category Of the action identification information belonging, generating conversion data including the category information indicating the other category to which the action identification information corresponding to the second action identification information belongs,
The information processing device according to claim 1 . The acquisition unit acquires a plurality of time-series data corresponding to each of a plurality of users,
A characteristic indicating behavior characteristics of the user, for each of the users corresponding to the time-series data acquired by the acquisition unit, by calculating an average of the plurality of behavior vectors generated based on the conversion data corresponding to the user. further comprising a user vector generation unit that generates a user vector that is a vector,
The information processing apparatus according to claim 1 or 2. further comprising a recommendation unit that recommends, to the user corresponding to the user vector, an action corresponding to the action vector relatively close to the user vector, among the plurality of action vectors generated by the action vector generation unit,
The information processing apparatus according to claim 3. The recommendation unit recommends the action corresponding to the user different from the user corresponding to the user vector,
The information processing apparatus according to claim 4. The time-series data is time information indicating the time when the user performed the action indicated by the action identification information, and further includes the time information associated with the action identification information,
The recommendation unit recommends an action indicated by the action identification information associated with the time information indicating a time included in a predetermined period in the time-series data,
The information processing apparatus according to claim 4 or 5. The time-series data includes action identification information that can identify the action position where the user performed the action,
The recommendation unit recommends an action indicated by the action identification information in which the action position is included in a predetermined range.
The information processing apparatus according to any one of claims 4 to 6. The recommendation unit recommends an action indicated by the action identification information that is not included in the time-series data corresponding to the user corresponding to the user vector,
The information processing apparatus according to any one of claims 4 to 7. the computer runs
Acquiring time-series data including action identification information for identifying the action for each action performed by the user and indicating the order in which the user performed each action;
From among a plurality of categories in which it is impossible to specify the action position where the user has acted, and to which each of the plurality of action identification information included in the time-series data belongs, the user selecting one said category to recommend to said user as a type of POI that may be of interest;
The action identification information belonging to the selected category, which is the category selected from among the plurality of action identification information included in the time-series data, is not converted, and is different from the selected category among the plurality of action identification information. a step of generating conversion data in which the action identification information belonging to the category of is converted into category information indicating the other category;
Based on the order of each action performed by the user indicated by the time-series data, for each of the action identification information or the category information included in the generated conversion data, the action indicated by the action identification information or the category information a step of generating an action vector, which is a feature vector indicating the relevance in order between actions belonging to and other actions different from the action;
A vector generation method with the computer,
an acquisition unit that acquires, for each action performed by a user, time-series data that includes action identification information for identifying the action and indicates the order in which the user performs each action;
From among a plurality of categories in which it is impossible to specify the action position where the user has acted, and to which each of the plurality of action identification information included in the time-series data belongs, the user a selection unit for selecting one of the categories to recommend to the user as a type of POI that may be of interest;
The action identification information belonging to the selected category which is the category selected by the selection unit among the plurality of action identification information included in the time-series data is not converted, and the selected category among the plurality of action identification information A data generation unit that generates conversion data obtained by converting the action identification information belonging to another category different from the category information to the category information indicating the other category, and
Based on the order of each action performed by the user indicated by the time-series data, the action identification information is indicated for each of the action identification information or the category information included in the conversion data generated by the data generation unit. an action vector generation unit that generates an action vector that is a feature vector that indicates the relationship in order between an action or an action belonging to the category information and another action that is different from the action;
A program to function as

Images