JP2018190154A - Recommendation system, device, and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To enable the recommendation of an application useful to a user without using a Web browsing history of the user.SOLUTION: A recommendation system relating to an embodiment includes: a use frequency storage part for storing the use frequency of each application of a plurality of users; a classification part for classifying a use tendency being a tendency of the use frequency of the user into a cluster; a recommendation part for recommending an application to a target user on the basis of the serviceability of each application in the cluster with the target user belonging thereto, and a recommendation condition; and a learning part for updating at least one of a classification method by the classification part and the recommendation condition on the basis of the use frequency stored in the use frequency storage part.SELECTED DRAWING: Figure 3

Description

  The present invention relates to a recommendation system, an apparatus, and a program.

  In recent years, an online market that provides various applications (hereinafter referred to as “apps”) to smart devices has been used. The user of the smart device can download a desired application from the market for a fee or free of charge and use it on the smart device. 2. Description of the Related Art Conventionally, a system that recommends an application useful for a user of a smart device is known.

  However, since the conventional system selects an app to be recommended based on the user's Web browsing history, it is difficult to select a useful application for the user when the Web browsing history cannot be acquired. .

  The present invention has been made in view of the above problems, and an object of the present invention is to make it possible to recommend an application that is useful to the user without using the user's Web browsing history.

  A recommendation system according to an embodiment includes a usage frequency storage unit that stores the usage frequency of each application of a plurality of users, a classification unit that classifies the usage trend of the usage frequency of the user into a cluster, and a target user Based on the usefulness of each application in the cluster to which the usage tendency belongs, and a recommendation condition, a recommendation unit that recommends the application to the target user, and the usage frequency stored in the usage frequency storage unit And a learning unit that updates at least one of the classification method by the classification unit and the recommendation condition.

  According to each embodiment of the present invention, an application useful for a user can be recommended without using the user's Web browsing history.

The figure which shows an example of a structure of a recommendation system. The figure which shows an example of the hardware constitutions of a user terminal. The figure which shows an example of a function structure of the recommendation system which concerns on 1st Embodiment. The figure which shows an example of an operation log. The figure which shows an example of usage frequency. The figure which shows an example of a usage tendency. The figure which shows an example of usefulness. The figure which shows an example of the usage tendency before and after installation of an application. The sequence diagram which shows an example of a use frequency and the update process of a cluster. The sequence diagram which shows an example of the recommendation process of an application. The sequence diagram which shows an example of the installation process of an application. The sequence diagram which shows an example of the update process of usefulness. The sequence diagram which shows an example of the learning process of a classification method and recommendation conditions. The figure which shows an example of a function structure of the recommendation system which concerns on 2nd Embodiment.

  Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings. In addition, regarding the description of the specification and the drawings according to each embodiment, constituent elements having substantially the same functional configuration are denoted by the same reference numerals and overlapping description is omitted.

<First Embodiment>
A recommendation system 1000 according to the first embodiment will be described with reference to FIGS. First, the configuration of the recommendation system 1000 will be described. FIG. 1 is a diagram illustrating an example of the configuration of the recommendation system 1000. A recommendation system 1000 in FIG. 1 includes user terminals 1 </ b> A to 1 </ b> B, a recommendation server 2, an application market 3, and a network 4.

  The user terminals 1A to 1C are computers for the user to use various applications. Although user terminal 1A-1C is PC (Personal Computer), a smart phone, a tablet terminal, a smart watch, and a television, for example, it is not restricted to this. Hereinafter, when the user terminals 1A to 1C are not distinguished from each other, they are simply referred to as user terminals 1. In the example of FIG. 1, the recommendation system 1000 includes three user terminals 1, but the number of user terminals 1 included in the recommendation system 1000 is arbitrary.

  The recommendation server 2 is an example of a recommendation device, and is a server computer that recommends an application useful for the user to the user of the user terminal. The recommendation server 2 recommends an application to each user terminal 1 user based on information acquired from each user terminal 1 included in the recommendation system 1000. In the present embodiment, the recommendation device may be, for example, a PC, a smartphone, a tablet terminal, or the like. Further, the recommendation device may be composed of a plurality of server computers.

  The application market 3 is an online market that provides users with paid or free applications, and is configured by one or a plurality of server computers. The user can download a desired application from the application market 3 to the user terminal.

  The network 4 includes a LAN (Local Area Network), the Internet, and the like, and connects the user terminal 1, the recommendation server 2, and the application market to each other.

  Next, the hardware configuration of the user terminal 1 according to the present embodiment will be described. FIG. 2 is a diagram illustrating an example of a hardware configuration of the user terminal 1. The user terminal 1 in FIG. 2 includes a CPU (Central Processing Unit) 101, a ROM (Read Only Memory) 102, a RAM (Random Access Memory) 103, and an HDD (Hard Disk Drive) 104. The user terminal 1 includes an input device 105, a display device 106, a communication interface 107, and a system bus 108.

  The CPU 101 executes the program to control each configuration of the user terminal 1 and realize the function of the user terminal 1.

  The ROM 102 stores a program (BIOS (Basic Input Output System) or the like) executed by the CPU 101 and various data.

  The RAM 103 provides a work area for the CPU 101.

  The HDD 104 stores programs (such as an OS (Operating System)) executed by the CPU 101 and various data.

  The input device 105 inputs information corresponding to a user operation to the user terminal 1. The input device 105 is, for example, a touch panel, a keyboard, a mouse, or a hardware button, but is not limited thereto.

  The display device 106 displays a screen corresponding to a user operation. The display device 106 is, for example, a liquid crystal display or an organic EL (Electro Luminescence) display, but is not limited thereto.

  The communication interface 107 is an interface for connecting the user terminal 1 to the network 4. The user terminal 1 communicates with the application market 3 and the recommendation server 4 connected to the network 4 via the communication interface 7.

  The system bus 108 connects the CPU 101, ROM 102, RAM 103, HDD 104, input device 105, display device 106, and communication interface 107 to each other.

  As with the user terminal 1, the recommendation server 2 includes a CPU 201, a ROM 202, a RAM 203, an HDD 204, an input device 205, a display device 206, and a communication interface 207. Connected to each other. Since each hardware is the same as that of the user terminal 1, description is abbreviate | omitted.

  Next, a functional configuration of the recommendation system 1000 according to the present embodiment will be described. FIG. 3 is a diagram illustrating an example of a functional configuration of the recommendation system 1000.

  First, the functional configuration of the user terminal 1 will be described. The user terminal 1 in FIG. 3 includes a terminal control unit 11, an application storage unit 12, an operation log storage unit 13, and an upload unit 14. Each of these functional configurations is realized by the CPU 101 executing a program and cooperating with each hardware of the user terminal 1.

  The terminal control unit 11 controls the overall operation of the user terminal 1. Specifically, the terminal control unit 11 downloads an application from the application market 3, installs and uninstalls the application, updates the operation log, and recommends the application to the user. The operation of the terminal control unit 11 will be described in detail later.

  The application storage unit 12 stores an application installed in the terminal control unit 11 and application information. The application information includes information such as an application name and an application ID. The application ID is application identification information.

  The operation log storage unit 13 stores an operation log of the user terminal 1 by the user. The operation log is data in which the operation execution date and time are associated with the operation content. The operation contents include downloading, installing, uninstalling, starting, and ending the application.

  FIG. 4 is a diagram illustrating an example of an operation log stored in the operation log storage unit 13. Each record in FIG. 4 corresponds to an operation log corresponding to each operation. In the example of FIG. 4, the download of the application A is executed at 15:00 on February 1, and the installation of the application A is executed at 15:00 on February 1.

  When the user terminal 1 is used by a plurality of users, the operation log corresponding to each operation may include the user ID of the user who executed the operation. The user ID is user identification information. Further, instead of the user ID being included in each operation log, the user's login may be included in the operation content stored in the operation log. In this case, the operation log corresponding to the user login includes the execution date and time (login date and time) and the user ID of the logged-in user. Hereinafter, a user whose user ID is X is referred to as user X.

  The upload unit 14 receives the operation log stored in the operation log storage unit 13 periodically or at a predetermined timing (timing when the operation log is updated, timing when the user terminal 1 is turned on, or the like). Upload to.

  Next, the functional configuration of the recommendation server 2 will be described. The recommendation server 2 in FIG. 3 includes a recommendation control unit 21, a usage frequency storage unit 22, a classification unit 23, a recommendation unit 24, and a learning unit 25. Each of these functional configurations is realized by the CPU 201 executing a program and cooperating with each hardware of the recommendation server 2.

  The recommendation control unit 21 controls the overall operation of the recommendation server 2. Specifically, the recommendation control unit 21 receives an operation log uploaded from the user terminal 1, updates the usage frequency, and requests an operation for each functional configuration. The operation of the recommendation control unit 21 will be described later in detail.

  The usage frequency storage unit 22 stores the usage frequency of each application of each user for each predetermined period. The use frequency may be the use count (start-up count) or the use period (start-up period). The predetermined period can be arbitrarily set such as one day, one week, or one month. Hereinafter, the usage frequency trend of each application is referred to as a usage trend.

  Further, the usage frequency storage unit 22 stores a cluster to which the usage tendency of each user for each predetermined period belongs. As will be described later, the usage tendency of each user for each predetermined period is classified into a plurality of clusters by the classification unit 23. The usage frequency storage unit 22 stores the clusters classified by the classification unit 23 in association with the usage tendency of each user for each predetermined period.

  FIG. 5 is a diagram illustrating an example of the usage frequency stored in the usage frequency storage unit 22. In the example of FIG. 5, the number of uses of the applications A to D for each week by the users U001 and U002 is stored in association with the cluster. For example, the usage count of the application A from 2/1 to 2/7 by the user U001 is 50 times, the usage count of the app B is 100 times, the usage count of the app C is 0 times, and the usage trend of the app of the user U001 Belongs to cluster 1. In FIG. 5, the value “−” of the usage count of the application indicates that the application is not installed. The application whose usage count is stored even though it is not installed corresponds to an application that is recommended but not installed, as will be described later.

  For example, the recommendation control unit 21 updates the usage frequency stored in the usage frequency storage unit 22 based on an operation log uploaded from each user terminal 1. Specifically, the recommendation control unit 21 counts the usage frequency of each application for each predetermined period based on the operation log uploaded from the user terminal 1, and calculates the usage frequency of each application of the user of the user terminal 1. Add to usage frequency. For example, when the usage frequency shown in FIG. 5 is stored in the usage frequency storage unit 22 and the operation log shown in FIG. 4 is uploaded from the user terminal 1, the recommendation control unit 21 starts 2 February from the user of the user terminal 1. What is necessary is just to add the usage frequency of application A and B of the period until the 6th of month one by one.

  The upload unit 14 of the user terminal 1 totals the usage frequency of each application for each predetermined period based on the operation log stored in the operation log storage unit 13 and uploads the totaled usage frequency to the recommendation server 2. May be. In this case, the recommendation server 2 may add the uploaded usage frequency to the usage frequency stored in the usage frequency storage unit 22 as it is.

  Based on the usage frequency stored in the usage frequency storage unit 22, the classification unit 23 periodically or at a predetermined timing (such as when the usage frequency is updated) Class. A cluster is a collection of similar usage trends. The clusters classified by the classification unit 23 are stored in the usage frequency storage unit 22 in association with the usage tendency (a set of usage frequencies of each application) belonging to the cluster.

  The classification unit 23 stores a classification method for classifying usage trends. The classification unit 23 can use an arbitrary clustering method such as the shortest distance method, the Ward method, or the k-average method as a classification method of the usage tendency. In this case, the classification unit 23 may calculate the distance between clusters (dissimilarity) based on the usage frequency value, or based on the usage frequency ratio.

  FIG. 6 is a diagram illustrating an example of a usage trend. In the example of FIG. 6, usage trends (a set of usage frequencies of the applications A to D) of the users U001 to U003 are shown. In the example of FIG. 6, the user U001 and the user U002 have similar usage frequency values, and the user U002 and the user U003 have similar usage frequency ratios.

  In the example of FIG. 6, when the classification unit 23 calculates the distance between the clusters based on the usage frequency value, the usage trends of the users U001 and U002 are classified into the same cluster, and the usage trends of the user U003 are different clusters. being classified. On the other hand, when the classification unit 23 calculates the distance between clusters based on the usage frequency ratio, the usage trends of the users U002 and U003 are classified into the same cluster, and the usage trends of the user U001 are classified into different clusters.

  Further, the classification unit 23 can use a method according to a predetermined classification condition as a classification method of the usage tendency. For example, the classification unit 23 may classify the usage trend for each application having the maximum usage frequency, or may classify the usage trend for each set of apps whose usage frequency or the ratio is equal to or greater than a threshold value. In the case of the former classification condition, in the example of FIG. 6, the usage trends of the users U002 and U003 are classified into the same cluster, and the usage trends of the user U001 are classified into different clusters.

  The recommendation unit 24 selects an application useful for the user regularly or at a predetermined timing (such as a use frequency or a timing when the cluster is updated), and recommends the selected application to the user. Specifically, the recommendation unit 24 transmits application information of an application selected as an application useful for the user to the user terminal 1 of the user.

  The recommendation part 24 memorize | stores the usefulness of each application for every cluster of a usage tendency. The usefulness of an application in a certain cluster is an index indicating the degree of usefulness of the application for a user whose usage tendency belongs to the cluster. The usefulness of an application in a certain cluster increases as the usage tendency of the application by a user belonging to the cluster increases. As will be described later, the usefulness is calculated by the learning unit 25. A method for calculating the usefulness will be described later.

  FIG. 7 is a diagram illustrating an example of the usefulness stored in the recommendation unit 24. In the example of FIG. 7, the usefulness of the applications A to D in the clusters 1 to 3 is shown. For example, the usefulness of the app A in the cluster 1 is 9, the usefulness of the app B is 7, the usefulness of the app C is 3, and the usefulness of the app D is 1. This means that the usage tendency of the application A is the highest for users who belong to the cluster 1 in the usage tendency.

  Moreover, the recommendation part 24 memorize | stores recommendation conditions. The recommendation condition corresponds to a selection condition for an application to be recommended to the user (useful application). As a recommendation condition, it is mentioned that usefulness is the maximum and usefulness is more than a threshold value.

  The recommendation unit 24 recommends, to a user belonging to a cluster having the usage tendency, an application that satisfies the recommendation level in the cluster as a useful application. For example, when the recommendation condition is that the usefulness is 7 or more, in the example of FIG. 7, the recommendation unit 24 recommends the apps A and B as useful apps to users whose usage tendency belongs to the cluster 1. .

  Even if the recommendation unit 24 is an application that satisfies the recommendation condition, it is preferable that the recommendation unit 24 does not recommend the application if the application is already installed in the user terminal 1. Thereby, the extra recommendation with respect to a user can be suppressed. The recommendation unit 24 can determine whether the application has been installed in the user terminal 1 by referring to the usage frequency of the user of the user terminal 1 stored in the usage frequency storage unit 22. Specifically, a field in which the usage frequency of the application is stored and an application frequency value is input (an app whose value is not “−”) corresponds to an installed app.

  Based on the usage frequency stored in the usage frequency storage unit 22, the learning unit 25 periodically or at a predetermined timing (such as a usage frequency or a cluster update timing) Update the conditions to increase the success probability of the recommendation. That is, the learning unit 25 learns a classification method and a recommendation condition with a high recommendation success probability.

  Updating the classification method means updating a clustering method parameter (such as k in the k-average method) and a classification condition parameter (such as a usage frequency threshold). Updating the recommendation condition means updating a parameter of the recommendation condition (such as a usefulness threshold). The success probability of the recommendation is a probability that the recommendation of the application by the recommendation unit 24 will be successful. The success of the recommendation means that the application recommended by the recommendation unit 24 is installed in the user terminal 1 and the recommended application is frequently used by the user of the user terminal 1 (the usage frequency satisfies a predetermined condition). Say. As the predetermined condition, for example, the usage frequency may be a predetermined value or more, and the increase rate of the usage frequency may be a predetermined value or more.

  In the present embodiment, the recommendation unit 24 recommends an application to the user, and the user installs the application in the user terminal 1 in accordance with the recommendation. In addition, the upload unit 14 transmits an operation log of the installed application to the recommendation server 2, and the usage frequency storage unit 22 stores the usage frequency of the application that is tabulated based on the operation log received from the user terminal 1. Further, the learning unit 25 learns the classification method used by the classification unit 23 and the recommendation conditions used by the recommendation unit 25 based on the application usage frequency stored in the usage frequency storage unit 22. As described above, the learning unit 25 is fed back the recommendation result of the application (whether or not the recommendation is successful), and learns the classification method and the recommendation condition based on the recommendation result. Therefore, the learning unit 25 can learn the classification method and the recommendation condition so that the success probability of the recommendation is increased.

  The learning unit 25 can learn a classification method and a recommendation condition with a high recommendation success probability using an arbitrary learning method such as a support vector machine (SVM), clustering, a neural network, or a decision tree. The learning unit 25 preferably includes a machine learning engine for learning a classification method and recommendation conditions.

  In addition, the learning unit 25 is useful for each application for each cluster based on the use frequency stored in the use frequency storage unit 22 periodically or at a predetermined timing (such as a use frequency or a timing when the cluster is updated). Update the degree. Here, the usefulness calculation method will be described using the usefulness of the app Y in the cluster X as an example.

  The learning unit 25 calculates the usefulness so that the usefulness of the app Y in the cluster X increases as the usage frequency of the app Y by the user whose usage tendency belongs to the cluster X increases. This is because an app that is frequently used by a user is considered to be a useful app for the user.

  In addition, when the app Y is installed in the user terminal 1 of the user after the app Y is recommended to the user whose usage tendency belongs to the cluster X, the learning unit 25 increases the usefulness of the app Y in the cluster X. As such, the usefulness may be calculated. This is because the application installed by the user is considered to be determined as a useful application by the user.

  In addition, when the app Y is uninstalled by a user whose usage tendency belongs to the cluster X, the learning unit 25 may calculate the usefulness so that the usefulness of the app Y in the cluster X decreases. This is because the application uninstalled by the user is considered to be determined as an unnecessary application by the user.

  In addition, after the app Y is installed by a user whose usage tendency belongs to the cluster X, the learning unit 25 has a high usefulness of the app Y in the cluster X when the cluster to which the usage trend of the user changes to the cluster Z. As such, the usefulness may be calculated. The cluster Z is a cluster different from the cluster X to which the usage tendency of users who frequently use the application Y belongs. This is because an application whose usage tendency changes from cluster X to cluster Z due to the installation is considered to be an extremely useful application for the user.

  Here, the change of the cluster will be specifically described. FIG. 8 is a diagram illustrating an example of a usage trend before and after installation of an application. The left side of FIG. 8 shows the usage tendency of the user before the application Y is installed. In the example of FIG. 8, before the application Y is installed, the applications A to D are installed in the user terminal 1, and the user usage tendency belongs to the cluster X.

  The right side of FIG. 8 shows the usage tendency of the user after the application Y is installed. When the installed application Y is not useful, as shown in the upper right of FIG. 8, the usage tendency after the installation of the application Y hardly changes, so the cluster to which the user usage tendency belongs does not change. On the other hand, when the installed application Y is useful, as shown in the lower right of FIG. 8, the user's usage tendency changes greatly, so the cluster to which the user's usage tendency belongs changes to cluster Z.

  The learning unit 25 can calculate the usefulness of the app Y in the cluster X based on the change from the cluster X to the cluster Z as described above by comparing the clusters before and after the installation of the app Y.

  Note that the cluster X may be changed to another cluster W when the app Y is not useful. However, since the cluster W is considered to be a cluster different from the cluster Z to which the usage tendency of the user whose usage frequency of the application Y is low belongs, the learning unit 25, when the cluster X changes to the cluster W, the cluster X It is preferable that the usefulness is not calculated so that the usefulness of the app Y is increased.

  It is also conceivable that the cluster X does not change when the app Y is useful. In this case, as described above, the usefulness is calculated so that the usefulness of the app Y in the cluster X is increased based on the usage frequency after the app Y is installed. The usefulness of can be calculated appropriately.

  Next, processing of the recommendation system 1000 according to the present embodiment will be described. FIG. 9 is a sequence diagram illustrating an example of usage frequency and cluster update processing.

  When the update timing arrives, the upload unit 14 of the user terminal 1 requests the operation log storage unit 13 for an operation log to be transmitted at the timing (step S101). Specifically, the upload unit 14 requests the operation log storage unit 13 for an operation log stored in the operation log storage unit 13 after the previous update timing (an operation log whose execution date and time is after the previous transmission timing). .

  When requested for an operation log, the operation log storage unit 13 passes the requested operation log to the upload unit 14 (step S102).

  Upon receiving the operation log, the upload unit 14 uploads the received operation log to the recommendation server 2 (step S103).

  When the upload unit 14 uploads the operation log, the recommendation control unit 21 receives the uploaded operation log, and totals the usage frequency of each application of each user for a predetermined period based on the operation log (step S104). .

  The recommendation control unit 21 sums up the usage frequencies, and stores the tabulated usage frequencies in the usage frequency storage unit 22 (step S105). Thereby, the usage frequency memorize | stored in the usage frequency memory | storage part 22 is updated.

  In addition, the recommendation control unit 21 requests the classification unit 23 to classify the usage tendency (step S106). In the example of FIG. 9, it is assumed that the classification unit 23 classifies the usage tendency according to the request from the recommendation control unit 21, but the classification unit 23 has a timing independent of the recommendation control unit 21. You can also classify usage trends.

  When the classification unit 23 is requested to classify the usage tendency, the classification unit 23 requests the usage frequency storage unit 22 for the usage frequency of each user's application (step S107). At this time, the classification unit 23 may request only the usage frequency of the period including the update timing (the usage frequency updated in step S107). Thereby, the load of the recommendation server 2 can be reduced.

  Further, the classification unit 23 may request the usage frequency of a plurality of periods including a period including the update timing. Thereby, the usage tendency of a plurality of periods including the period including the update timing can be classified by the same classification method. As a result, even if the classification method is updated by the learning unit 25 between the previous update timing and the current update timing, the learning unit 25 more appropriately selects each application based on the change in the cluster. The usefulness of can be calculated. In other words, when the usage trends of a plurality of periods including a period including the update timing are classified by different classification methods, the learning unit 25 may not be able to appropriately calculate the usefulness of each application based on the change in the cluster. There is. This is because it is not known whether the change in the cluster is due to the application installation or the change in the classification method.

  When the usage frequency is requested, the usage frequency storage unit 22 passes the requested usage frequency to the classification unit 23 (step S108).

  Upon receiving the usage frequency, the classification unit 23 classifies the usage tendency of each user for each predetermined period based on the received usage frequency (step S109). The usage trend classification method is as described above.

  When classifying the usage trend, the classification unit 23 stores the cluster obtained by the classification in the usage frequency storage unit 22 (step S110). Thereby, the cluster memorize | stored in the utilization frequency memory | storage part 22 is updated.

  Through the above processing, the recommendation server 2 can update the usage frequency and the cluster based on the latest operation log. As a result, the usage frequency and cluster as shown in FIG. 5 are stored in the usage frequency storage unit 22.

  FIG. 10 is a sequence diagram illustrating an example of an application recommendation process. Hereinafter, a user who is a recommendation target of an application is referred to as a target user. There may be one or more target users.

  When the recommendation timing arrives, the recommendation control unit 2 of the recommendation server 2 requests the recommendation unit 24 to recommend the app to the target user (step S201). In the example of FIG. 10, it is assumed that the recommendation unit 24 recommends an application in response to a request from the recommendation control unit 21, but the recommendation unit 24 executes the application at a timing independent of the recommendation control unit 21. You may make recommendations.

  When the recommendation unit 24 is requested to recommend an application, the recommendation unit 24 requests the usage frequency storage unit 22 for a cluster to which the usage frequency of the target user belongs in a period including the recommendation timing (step S202).

  When requested to use a cluster, the usage frequency storage unit 22 passes the requested cluster to the recommendation unit 24 (step S203).

  When receiving the cluster, the recommendation unit 24 refers to the usefulness and recommendation conditions of the received cluster (target user's cluster), and selects an application whose usefulness satisfies the recommendation condition as a useful application (step S204).

  When the recommendation unit 24 selects an application, the recommendation unit 24 transmits application information of the selected application to the user terminal 1 of the target user (step S206). The application information may include a description of the application, the URL of the application in the application market 3, and the like.

  When the recommendation unit 24 transmits the application information, the terminal control unit 11 of the user terminal 1 of the target user receives the application information and stores the operation log in the operation log storage unit 13 (step S206). The operation content of the operation log stored here is an application recommendation.

  Further, the terminal control unit 11 recommends a useful application selected by the recommendation unit 24 to the target user based on the received application information (step S207). The recommendation method of the application by the terminal control part 11 is arbitrary. For example, the terminal control unit 11 may recommend an application by displaying a popup or a dialog on the display device 106 of the user terminal 1, or may recommend the application by push notification or email transmission.

  On the other hand, after transmitting the application information to the user terminal 1, the recommendation unit 24 adds the field of the selected application to the usage frequency of the target user in the usage frequency storage unit 13 (step S208). At this time, the usage frequency of the application is set to “−” as described above (see application D in FIG. 5). If the field of the selected application already exists, the recommendation unit 24 does not have to add a field.

  Through the above processing, the recommendation server 2 can recommend an application useful for the target user. Further, the usage frequency and cluster as shown in FIG. 5 are stored in the usage frequency storage unit 22.

  FIG. 11 is a sequence diagram illustrating an example of an application installation process. Hereinafter, an application installed in the user terminal 1 is referred to as a target application.

  When the user of the user terminal 1 operates the input device 105 to instruct the user terminal 1 to download the target application, the terminal control unit 11 requests the application market 3 to download the target application (step S301).

  When requested to download the target application, the application market 3 transmits the requested target application and application information of the target application to the user terminal 1 (step S302).

  Upon receiving the target application and application information downloaded from the application market 3, the terminal control unit 11 stores the received target application and application information in the application storage unit 12 (step S303).

  Further, the terminal control unit 11 stores an operation log in the operation log storage unit 13 (step S304). The operation content of the operation log stored here is download of the target application.

  Thereafter, when the user of the user terminal 1 operates the input device 105 to instruct the user terminal 1 to install the target application, the terminal control unit 11 requests the application storage unit 12 for an installer of the target application (step S305). .

  When requested to install the target application, the application storage unit 12 passes the requested installer to the terminal control unit 11 (step S306).

  When receiving the installer of the target application, the terminal control unit 11 executes the received installer and installs the target application in the user terminal 1 (step S307).

  When the installation of the target application is completed, the terminal control unit 11 stores the setting information of the target application generated by the installation in the application storage unit 12 (Step S308).

  In addition, the terminal control unit 11 stores an operation log in the operation log storage unit 13 (step S309). The operation content of the operation log stored here is installation of the target application.

  Thereafter, the user can use the target application at a desired timing. Each time the user uses the target application, an operation log corresponding to the usage content is stored in the operation log storage unit 13.

  Through the above processing, the user can install the target application in the user terminal 1. Also, an operation log as shown in FIG. 4 is stored in the operation log storage unit 13.

  FIG. 12 is a sequence diagram illustrating an example of a usefulness update process.

  When the usefulness update timing comes, the recommendation control unit 21 of the recommendation server 2 requests the learning unit 25 to update the usefulness level (step S401). In the example of FIG. 12, it is assumed that the learning unit 25 updates the usefulness in response to a request from the recommendation control unit 21, but the learning unit 25 is useful at a timing independent of the recommendation control unit 21. You may update the degree.

  When requested to update the usefulness, the learning unit 25 requests the usage frequency and cluster from the usage frequency storage unit 22 (step S402). The learning unit 25 may request a usage frequency and a cluster corresponding to the usefulness calculation method. For example, the learning unit 25 requests a use frequency and a cluster for a period including the update timing. Thereby, the learning unit 25 can calculate the usefulness based on the usage frequency. In addition, the learning unit 25 may request the usage frequency and the cluster of a plurality of periods including a period including the update timing. Thereby, the learning unit 25 can calculate the usefulness based on the usage frequency, application installation, application uninstallation, and cluster change.

  When the usage frequency and cluster are requested, the usage frequency storage unit 22 passes the requested usage frequency and cluster to the learning unit 25 (step S403).

  Upon receiving the usage frequency and the cluster, the learning unit 25 calculates the usefulness of each application for each cluster based on the received usage frequency and the cluster (step S404). The calculation method of the usefulness is as described above.

  When the learning unit 25 calculates the usefulness level, the learning unit 25 passes the calculated usefulness level to the recommendation unit 24 (step S405). When the recommendation unit 24 receives the usefulness, the recommendation unit 24 stores the received usefulness and deletes the original usefulness. Thereby, the usefulness which the recommendation part 24 memorize | stores is updated. Thereafter, the recommendation unit 24 selects a useful application using the updated usefulness.

  The usefulness can be updated by the above processing. By updating the usefulness as needed, the recommendation unit 24 can recommend a useful application according to the latest usage frequency to the user.

  FIG. 13 is a sequence diagram illustrating an example of a classification method and recommendation condition learning processing.

  The recommendation control unit 21 of the recommendation server 2 requests the learning unit 25 to learn the classification method and the recommendation condition when the learning timing of the classification method and the recommendation condition arrives (step S501). In the example of FIG. 13, it is assumed that the learning unit 25 learns the classification method and the recommendation condition according to the request from the recommendation control unit 21, but the learning unit 25 is independent of the recommendation control unit 21. You may learn a classification method and recommendation conditions with timing.

  When learning unit 25 is requested to learn the classification method and recommendation conditions, learning unit 25 requests use frequency and cluster from use frequency storage unit 22 (step S502). The learning unit 25 may request a usage frequency and a cluster corresponding to the classification method and the learning method of the recommendation condition.

  When the usage frequency and cluster are requested, the usage frequency storage unit 22 passes the requested usage frequency and cluster to the learning unit 25 (step S503).

  Upon receiving the usage frequency and the cluster, the learning unit 25 learns a classification method and a recommendation condition with a high recommendation success probability based on the received usage frequency and cluster (step S504). The classification method and the recommendation condition learning method are as described above.

  When learning unit 25 learns the classification method, learning unit 25 passes the learned classification method (parameter) to classification unit 23 (step S505). Upon receiving the classification method, the classification unit 23 stores the received classification method and deletes the original classification method. Thereby, the classification method stored in the classification unit 23 is updated. Thereafter, the classification unit 23 classifies the usage tendency using the updated classification method.

  In addition, when learning unit 25 learns the recommendation condition, learning unit 25 passes the learned recommendation condition (parameter) to recommendation unit 24 (step S506). When receiving the recommendation condition, the recommendation unit 24 stores the received recommendation condition and deletes the original recommendation condition. Thereby, the recommendation condition which the recommendation part 24 memorize | stores is updated. Thereafter, the recommendation unit 24 recommends useful applications using the updated recommendation conditions.

  With the above processing, the classification method and the recommendation condition can be updated. By updating the classification method as needed, the classification unit 23 can classify the user's usage tendency by a classification method according to the latest usage frequency. Moreover, the recommendation part 24 can recommend a useful application by the recommendation conditions according to the newest usage frequency by updating recommendation conditions at any time.

  As described above, the recommendation system 1000 according to the present embodiment recommends applications that are useful to the target user based on the usage tendency of the applications of the plurality of users including the target user. The usage tendency of the application can be acquired by counting the operation logs received from the user terminal 1. Therefore, the recommendation system 1000 can recommend an application useful for the target user without using the Web browsing history of the target user.

  In the present embodiment, the recommendation system 1000 can also recommend an application that is useful to the target user by using both the application usage trend of the target user and the Web browsing history of the target user.

Second Embodiment
A recommendation system 1000 according to the second embodiment will be described with reference to FIG. FIG. 14 is a diagram illustrating an example of a functional configuration of the recommendation system 1000 according to the present embodiment. A recommendation system 1000 in FIG. 14 includes a user terminal 1 and an application market 3.

  The user terminal 1 includes a terminal control unit 11, an application storage unit 12, an operation log storage unit 13, a recommendation control unit 21, a usage frequency storage unit 22, a classification unit 23, a recommendation unit 24, and a learning unit 25. And comprising. The function of each functional configuration is the same as in the first embodiment. That is, in this embodiment, the user terminal 1 is an example of a recommendation device, and includes the function of the recommendation server 2 in the first embodiment. The recommendation control unit 21 of the user terminal 1 receives operation logs from other user terminals included in the recommendation system 1000, totals the usage frequencies based on the received operation logs, and uses the tabulated usage frequencies as the usage frequency storage unit 22. save.

  With the configuration as described above, the recommendation system 1000 according to the present embodiment can recommend an application useful to the target user without using the Web browsing history of the target user, as in the first embodiment.

  Note that, in the present embodiment, the recommendation system 1000 recommends a useful application for the target user by using both the usage trend of the target user's application and the target user's Web browsing history, as in the first embodiment. It is also possible to do.

  In the present embodiment, all user terminals included in the recommendation system 1000 may each have the function of the recommendation server 2, or some user terminals have the function of the recommendation server 2, and other user terminals have You may recommend an app.

  Each function of the embodiment described above can be realized by one or a plurality of processing circuits. Here, the “processing circuit” in this specification means a processor programmed to execute each function by software, such as a processor implemented by an electronic circuit, or designed to execute each function described above. It also includes devices such as application specific integrated circuits (ASICs), digital signal processors (DSPs), field programmable gate arrays (FPGAs), and conventional circuit modules.

  It should be noted that the present invention is not limited to the configuration shown here, such as a combination with other elements in the configuration described in the above embodiment. These points can be changed without departing from the spirit of the present invention, and can be appropriately determined according to the application form.

1: User terminal 2: Recommendation server 3: Application market 11: Terminal control unit 12: Application storage unit 13: Operation log storage unit 14: Upload unit 21: Recommendation control unit 22: Usage frequency storage unit 23: Classification unit 24: Recommendation Part 25: Learning part 1000: Recommendation system

JP 2015-22507 A

Claims (11)

A usage frequency storage unit for storing the usage frequency of each application of a plurality of users;
A classifying unit that classifies the usage tendency, which is the tendency of the usage frequency of the user, into clusters;
A recommendation unit that recommends an application to the target user based on the usefulness of each application in the cluster to which the usage tendency of the target user belongs and a recommendation condition;
Based on the usage frequency stored in the usage frequency storage unit, a learning unit that updates at least one of the classification method by the classification unit and the recommendation condition;
A recommendation system with The recommendation system according to claim 1, wherein the classification unit calculates a distance between the clusters based on a value or a ratio of the usage frequency. The recommendation system according to claim 1 or 2, wherein the learning unit updates at least one of the classification method and the recommendation condition so that a success probability of the recommendation is increased. 4. The recommendation system according to claim 3, wherein the success of the recommendation includes that the usage frequency of the application by the target user satisfies a predetermined condition after the recommended application is installed by the target user. The recommendation system according to any one of claims 1 to 4, wherein the recommendation unit recommends, to the target user, an application whose usefulness satisfies the recommendation condition. The recommendation system according to claim 1, wherein the usefulness of the application is calculated to be higher as the usage frequency of the application is higher. The recommendation system according to any one of claims 1 to 6, wherein the usefulness of the application is calculated to be high when the application is installed by the user. The recommendation system according to any one of claims 1 to 7, wherein the usefulness of the application is calculated to be low when the application is uninstalled by the user. The recommendation according to any one of claims 1 to 8, wherein the usefulness of the application is calculated to be high when the cluster to which the usage tendency of the user belongs changes due to installation of the application. system. A usage frequency storage unit for storing the usage frequency of each application of a plurality of users;
A classifying unit that classifies the usage tendency, which is the tendency of the usage frequency of the user, into clusters;
A recommendation unit that recommends an application to the target user based on the usefulness of each application in the cluster to which the usage tendency of the target user belongs and a recommendation condition;
Based on the usage frequency stored in the usage frequency storage unit, a learning unit that updates at least one of the classification method by the classification unit and the recommendation condition;
A recommendation device comprising: A usage frequency storage step for storing the usage frequency of each application of a plurality of users;
A classification step of classifying the usage trend, which is the trend of the usage frequency of the user, into clusters;
A recommendation step of recommending an application to the target user based on the usefulness of each application in the cluster to which the usage tendency of the target user belongs and a recommendation condition;
A learning step of updating at least one of the classification method and the recommendation condition in the classification step based on the usage frequency stored in the usage frequency storage step;
A program that causes a computer to execute.

Images