JP7145901B2 - Information processing device, information processing method and information processing program - Google Patents

Description

The present invention relates to an information processing device, an information processing method, and an information processing program.

Techniques for grasping the actual state of withdrawal of users from online services have been disclosed. For example, the content transition of the online service used by the user is detected, and from the degree of transition to other content for each content, the transition eigenvector, which is the probability of staying in the state, is calculated. A technique is known in which a degree of not transitioning to a content is calculated for each content as a probability of leaving, and a value obtained by multiplying a transition eigenvector for each content by the probability of leaving for each content is calculated as a leaving factor value.

JP 2019-152991 A

However, it may not be possible to say that the above conventional technology accurately estimates the user's withdrawal mode. For example, the above-described conventional technology merely estimates the user's withdrawal mode based on the withdrawal probability for each content. For this reason, it cannot be said that the intention (context) of the user's withdrawal is properly estimated.

The present application has been made in view of the above, and aims to find out the cause of withdrawal more effectively.

An information processing apparatus according to the present application includes an acquisition unit that acquires sequence data indicating the transition of use of a service by each user, the sequence data including use of a predetermined target service; and a learning unit that causes a model to learn to output withdrawal period information indicating a period from when the target service is used until when the target service is used again.

According to one aspect of the embodiment, the cause of withdrawal can be found more effectively.

FIG. 1 is an explanatory diagram showing an outline of an information processing method according to an embodiment. FIG. 2A is an explanatory diagram showing an outline of a purpose-of-use prediction model. FIG. 2B is an explanatory diagram showing an overview of the withdrawal period prediction model. FIG. 2C is an explanatory diagram showing a transition sequence to each function of one application. FIG. 3 is a diagram illustrating a configuration example of an information processing system according to the embodiment; FIG. 4 is a diagram illustrating a configuration example of an information processing apparatus according to the embodiment; FIG. 5 is a diagram showing an example of a user database. FIG. 6 is a diagram showing an example of a sequence database. FIG. 7 is a diagram showing an example of a withdrawal period database. FIG. 8 is a flow chart showing a processing procedure according to the embodiment. FIG. 9 is a diagram illustrating an example of a hardware configuration;

Hereinafter, modes for implementing an information processing apparatus, an information processing method, and an information processing program according to the present application (hereinafter referred to as "embodiments") will be described in detail with reference to the drawings. The information processing apparatus, information processing method, and information processing program according to the present application are not limited to this embodiment. Also, in the following embodiments, the same parts are denoted by the same reference numerals, and overlapping descriptions are omitted.

[1. Outline of information processing method]
First, an outline of an information processing method performed by an information processing apparatus according to an embodiment will be described with reference to FIG. FIG. 1 is an explanatory diagram showing an outline of an information processing method according to an embodiment. Note that in FIG. 1, a case will be described as an example in which the reason for leaving the target service (the reason why the user stopped using the service) is estimated by machine learning.

As shown in FIG. 1 , the information processing system 1 includes a terminal device 10 , an information processing device 100 and an information providing device 200 . The terminal device 10, the information processing device 100, and the information providing device 200 are communicably connected to each other by wire or wirelessly via a network N (see FIG. 3).

The terminal device 10 is a smart device such as a smartphone or tablet used by the user U, and communicates with an arbitrary server device via a wireless communication network such as 4G (Generation) or LTE (Long Term Evolution). It is an information processing device that can The terminal device 10 has a screen such as a liquid crystal display, which has a touch panel function. Accepts various operations. An operation performed on an area where content is displayed on the screen may be an operation on the content. In addition, the terminal device 10 may be an information processing device such as a desktop PC (Personal Computer) or a notebook PC as well as a smart device.

The information providing device 200 is an information processing device that provides various services to the terminal device 10 of each user U, and is realized by a server device, a cloud system, or the like. A plurality of information providing devices 200 may be provided. In addition, the information providing device 200 provides a function of searching for content such as web pages, websites, and/or image files existing on the Internet in response to a search request from the terminal device 10 of the user U. It may be a server (search engine). For example, the information providing apparatus 200 may provide search sites including portal sites and the like. Here, when the terminal device 10 or the like accesses the information providing device 200, the information providing device 200 manages the state between the information providing device 200 and the web browser from the terminal device 10 or the like via the web browser. A communication protocol or data stored in a web browser used therein, such as an HTTP cookie, may be obtained. Also, the information providing apparatus 200 may be a relay server (Proxy), a redirector (Redirector), or the like.

The information processing device 100 is an information processing device that analyzes the cause of withdrawal of each user U, and is realized by a server device, a cloud system, or the like. Note that the information processing device 100 may be one of the information providing devices 200 . That is, the information processing apparatus 100 may also be an information processing apparatus that provides services.

For example, the user U uses the terminal device 10 to use various services provided by each information providing device 200 (step S1).

Each information providing device 200 aggregates accesses from the terminal device 10 of each user U and generates an access log for each user U (step S2).

The information processing device 100 acquires the access log of each user U from each information providing device 200 (step S3).

At this time, the information processing apparatus 100 may acquire only the access log of the user U who has given prior consent to the use of the access log. For example, the information processing apparatus 100 may acquire only the access log of the user U registered in advance.

When the information processing apparatus 100 itself provides services to the terminal devices 10 of the users U, the information processing apparatus 100 can generate an access log of each user U regarding the services that the information processing apparatus 100 provides. .

The information processing apparatus 100 creates service use transition sequence data regarding service use in the same session from the access log of each user U (step S4).

Incidentally, in practice, a series of accesses by the user U within a predetermined period may be regarded as use of the service in the same session. In addition to the access made within a predetermined period, for example, when following a link, the service may be used in the same session.

In addition to access from the same browser or from the same application, access from different browsers or different applications in the same time zone may be used in the same session. That is, in addition to continuous accesses from the same browser or the like, continuous accesses within a predetermined period may be regarded as use of the service in the same session.

Also, the service may be any service. For example, the service may be web content or an application (hereinafter referred to as an application). Further, the service may be a program function, data, or the like provided to the terminal device 10 via an API (Application Programming Interface). Also, the unit of service may be per web content, per application, per domain, per subdomain, or the like. Further, the information processing apparatus 100 may treat the behavior of the user to be analyzed as a service with arbitrary granularity, and create service usage transition sequence data indicating transition for each service.

Here, the information processing apparatus 100 records an access log with arbitrary granularity, such as the order of domains accessed by the user U, the order of applications accessed by the user U, or the order of web content accessed by the user U, for example. In summary, service use transition sequence data indicating the contents of services used by the user U and their order is created.

In the example shown in FIG. 1, this service usage transition sequence data includes "first search site", "local summary site", "first map WEB", "first map application", "first search site", " It includes "travel site" and "accommodation reservation site", and indicates that the use of services is transitioning in the order of these.

"First search site" indicates a search service (search engine) provided via a web browser. A "local summary site" indicates a summary site of local information, sightseeing information, and the like accessed by the user U using the web browser. "First map WEB" indicates a map service provided via a web browser. "First map application" indicates a map service provided via a dedicated map application. "Travel site" indicates a travel-related site accessed by the user U using the web browser. "Accommodation reservation site" indicates a reservation site such as an accommodation facility accessed by the user U using the web browser.

In generating such service usage transition sequence data, a user ID that is commonly used by the user U across services may be used. For example, a table may be used that converts user IDs that are individually used by users U to common user IDs in each service. Further, this user ID may be identification information of the terminal device 10, for example.

The information processing apparatus 100 inputs the service usage transition sequence data into the model, learns the intention (usage purpose) of the user U using the service in the same session, and calculates the withdrawal period (absence period) based on the learning result. Time) is created (step S5). In practice, a model for predicting the withdrawal period may be learned from the beginning without learning the purpose of use.

At this time, the information processing apparatus 100 causes the model to learn by deep learning using the withdrawal period as correct data. For example, the sequence data indicates the services used by the user U and the order in which the user U used them. Therefore, the information processing apparatus 100 learns a model so as to output a withdrawal period when a service and an order are input. Then, the information processing apparatus 100 generates and outputs the predicted withdrawal period based on the model for predicting the withdrawal period.

The information processing apparatus 100 sets a certain percentage of the learning data as test data in advance, inputs all the service usage transition sequence data of the test data into the model, and predicts a long withdrawal period. They are rearranged in order, and the reason for leaving the target service is estimated (step S6). Note that the withdrawal period may be predicted by inputting not only test data but also learning data.

Here, the information processing apparatus 100 defines an interval (interval) from the end of use of the target service to the start of use of the next target service as the withdrawal period. Note that the withdrawal period is not limited to time, and may be, for example, the number of services (number of contents) used by the user U from the time of using the most recent target service to the time of using the next target service. That is, the withdrawal period is withdrawal period information indicating the period from when the user U uses the target service until when he/she uses the target service again.

Also, considering the session, the usage sequence may be from the time of using the most recent target service to the time of using the target service in the next session. That is, the service use transition sequence data may be created with one cycle from the time of use of the most recent target service to the time of use of the target service in the next session.

Note that the withdrawal period continues to increase unless the user U uses the service again. If the user U never uses the service again, the withdrawal period approaches infinity. In this case, an upper limit may be set for the withdrawal period. For example, when the withdrawal period reaches a predetermined threshold, the information processing apparatus 100 fixes the withdrawal period to the threshold, and determines that the user U has completely withdrawn from the service (no longer uses it). You may do so. Alternatively, a binary classification may be made as to whether the user U will not use the service next time or otherwise.

Also, if there is a period during which the service cannot be used physically, regardless of the circumstances on the user U side, such as service failure or maintenance, that period may be excluded from the withdrawal period. However, frequent occurrences of such troubles, maintenance, etc. may cause the user U to become dissatisfied and stop using the service, so it may be positively included in the withdrawal period. In this case, for example, as the breakdown of the withdrawal period, the withdrawal period due to factors on the user U side and the withdrawal period due to factors on the service side (malfunction, maintenance, etc.) may be indicated.

Also, the withdrawal period is an index indicating whether or not the user U is satisfied with/attracted to the service. For example, if the withdrawal period is short relatively/compared to a predetermined threshold, it indicates that the user U is actively using the service and that the user U is satisfied/attractive with the service. It is assumed that he feels Conversely, if the withdrawal period is long, it is presumed that the user U does not use the service very much and that the user U is dissatisfied with/not attracted to the service. In other words, the withdrawal period can also be index information that indicates the user's impression of the target service.

Therefore, the information processing apparatus 100 can determine that the reason for leaving the target service is included in the service usage transition sequence data with the long leaving period of the target service. By examining the service usage transition sequence data, it is possible to estimate the reason for leaving the target service and improve the service.

Here, regarding the service usage transition sequence data, the sequence pattern and the span of the withdrawal period are different between services that are used on a daily basis and services that are used at specific times of the year such as Obon/New Year/Golden Week (GW). different.

Also, even with the same map application, when it is used for weather forecast confirmation such as rain cloud radar, and when it is used for traffic information confirmation such as congestion forecast during consecutive holidays, the sequence pattern and departure period span are different. is different.

Therefore, in the present embodiment, the purpose of use for each sequence is specified first, and then a model for predicting the withdrawal period of the target service is created.

[1-1. Usage purpose prediction model]
The usage purpose prediction model will be described with reference to FIG. 2A. FIG. 2A is an explanatory diagram showing an outline of a purpose-of-use prediction model.

In the present embodiment, the information processing apparatus 100 inputs service usage transition sequence data to the model and makes the model learn about the usage purpose of the sequence.

Any type of model can be adopted as the model. For example, the information processing apparatus 100 may employ an SVM (Support Vector Machine) or a DNN (Deep Neural Network) as a model. Here, the DNN may be a CNN (Convolutional Neural Network) or an RNN (Recurrent Neural Network). Also, the RNN may be LSTM (Long short-term memory) or the like. That is, any model can be adopted as the model. The model may also be a model realized by combining a plurality of models, such as a model combining CNN and RNN.

In addition, the information processing apparatus 100 regards the behavior of the user to be analyzed as a service with arbitrary granularity, creates service usage transition sequence data indicating transition for each service, and creates service usage transition sequence data indicating features (service A model may be generated from the characteristics of the usage transition mode).

In this embodiment, the case of using LSTM as an example of a model will be described. LSTM is a model for sequential data.

First, the information processing apparatus 100 inputs the service use transition sequence data to the LSTM, and makes the model learn to predict the service to be used next to each service included in the service use transition sequence data. . That is, the model learns to find a service that has a high probability of being used next to that service.

Here, when the service "first search site" is input to the LSTM, the information processing apparatus 100 makes the model learn to predict the service "local summary site" as the next service to be used after the service.

This prediction is presumed to be correct because the service 'local summary site' is actually used after the service 'first search site' in the input service use transition sequence data.

The information processing apparatus 100 uses the same procedure to make the model learn to predict the next service to be used after the service "local summary site".

Thus, in the example shown in FIG. 2A, "first search site", "local summary site", "first map web". . . Enter the "value indicating each service" in the order in which each service was used. At this time, the information processing apparatus 100 performs pre-training (pre-learning) of the model so as to predict the service following the input service. For example, the model is pre-trained to output "local summary site" when "first search site" is input. In addition, it is not necessary to actually perform pre-training.

Such a model will eventually learn the user's impression (purpose of use, satisfaction or not, degree of satisfaction, etc.) when using the target service in the sequence data.

It should be noted that a service whose purpose of use is similar to each service may be obtained by separately quantifying the purpose of use of each service by the information processing apparatus 100 .

For example, the information processing apparatus 100 uses a learning method such as a vector representation method (for example, Word2Vec) or a similar method to make the model learn the features of the service usage transition sequence data. Then, the information processing apparatus 100 uses the learned model to generate a vector for each service. The information processing apparatus 100 visualizes and analyzes the vector for each service, thereby estimating a service whose purpose of use is close to each service.

In addition, a distributed expression (word embedding) indicating the purpose of use of the sequence is generated at the top right end of the LSTM model shown in FIG. 2A. Distributed representation expresses a word by a high-dimensional (approximately 200-dimensional) real number vector. Distributed representations associate words with close meanings with close vectors. Here, the distributed representation expresses the purpose of use of the sequence.

[1-2. Withdrawal period prediction model]
The withdrawal period prediction model will be described with reference to FIG. 2B. FIG. 2B is an explanatory diagram showing an overview of the withdrawal period prediction model. Note that the withdrawal period prediction model may be learned from the beginning without learning the purpose of use prediction model.

The information processing apparatus 100 inputs all the service usage transition sequence data learned by the usage purpose prediction model into the model. The model is for example LSTM. However, LSTM is only an example.

Thereby, the information processing apparatus 100 inputs the service usage transition sequence data to the model and calculates the squared error. Then, the information processing apparatus 100 performs model learning so as to reduce the squared error.

The squared error is the error between the actual value and the model-predicted value. That is, it indicates the deviation between the actual withdrawal duration and the withdrawal duration predicted by the model.

The information processing apparatus 100 sets a certain percentage of the learning data as test data in advance, inputs all the service usage transition sequence data of the test data into the model, and predicts a long withdrawal period. Sort by order. Moreover, the data used for learning may be used as input data. For example, the information processing apparatus 100 outputs the service use transition sequence data in descending order of the withdrawal period. At this time, the information processing apparatus 100 may limit the number of service usage transition sequence data to be output to a predetermined number. Further, the information processing apparatus 100 sorts (arranges) all the input service usage transition sequence data in ascending/descending order and merges duplicate data based on the predicted withdrawal period. You can Further, the information processing apparatus 100 may compare service use transition sequence data output in order from the longest withdrawal period with service use transition sequence data output in order from the shortest withdrawal period. This reveals the characteristics of a transition sequence group with a long departure period (high probability of departure), that is, a transition sequence group including a cause of departure. As a result, it is possible to interpret this transition sequence group as having a cause for leaving, and to improve the service.

As described above, in the example shown in FIG. 2B, the information processing apparatus 100 outputs a Calculate the squared error between the predicted value and the actual value, and re-learn the model so that the squared error becomes smaller. In addition, it is not necessary to actually perform pre-training.

As a result, the information processing apparatus 100 can make a model for estimating the withdrawal period learn from the sequence data.

[1-3. An example of a fine-grained service usage transition sequence]
So far, we have explained the case of inputting coarse-grained service usage transition sequence data, such as transitions to different sites. The case can be similarly processed.

With reference to FIG. 2C, a case where fine-grained service usage transition sequence data is input will be described. FIG. 2C is an explanatory diagram showing a transition sequence to each function of one application.

As shown in FIG. 2C, the transition sequence to each function in the map service "first map" is limited to those in which the "gourmet icon" is included in the sequence, and the learned sequence pattern is input. .

In the example shown in FIG. 2C, for the map service "first map", this sequence pattern starts with "first map application launch", "map search", "result screen scroll", "gourmet icon", "result screen scroll”, “gourmet detail display” and “telephone”.

By doing so, it is possible to estimate the purpose of use of each function of one application and find out the cause of withdrawal in the transition sequence of each function. For example, if the sequence pattern with a long withdrawal period is characterized by the use of a certain function, it may be that the function is becoming difficult to use as the times change due to poor design of the application, or the user U It is presumed that he became dissatisfied with the function. This clarifies issues for service improvement.

In this embodiment, the reason for leaving the target service is estimated by deep learning. This allows the model to derive the sequence of usages that the user leaves the most.

[2. Configuration example of information processing system]
Next, the configuration of the information processing system 1 including the information processing device 100 according to the embodiment will be described with reference to FIG. FIG. 3 is a diagram showing a configuration example of the information processing system 1 according to the embodiment. As shown in FIG. 3, the information processing system 1 according to the embodiment includes a terminal device 10, an information processing device 100, and an information providing device 200. FIG. These various devices are communicatively connected via a network N by wire or wirelessly. The network N is, for example, a LAN (Local Area Network) or a WAN (Wide Area Network) such as the Internet.

Also, the number of devices included in the information processing system 1 shown in FIG. 3 is not limited to the illustrated one. For example, in FIG. 3, only one terminal device 10 and one information providing device 200 are shown for simplification of illustration, but this is only an example and is not limited, and two or more devices are possible. good too.

The terminal device 10 is an information processing device used by the user U. FIG. For example, the terminal device 10 includes smart devices such as smartphones and tablet terminals, feature phones, PCs (Personal Computers), PDAs (Personal Digital Assistants), car navigation systems, and wearable devices such as smart watches and head-mounted displays. , smart glasses, etc.

In addition, the terminal device 10 is compatible with wireless communication networks such as LTE (Long Term Evolution), 4G (4th Generation), 5G (5th Generation: fifth generation mobile communication system), Bluetooth (registered trademark), wireless LAN (Local It is possible to connect to the network N via short-range wireless communication such as Area Network) and communicate with the information processing device 100 and the information providing device 200 .

The information processing device 100 and the information providing device 200 are, for example, a PC, a server, a mainframe, a workstation, or the like. Note that the information processing device 100 and the information providing device 200 may be realized by cloud computing.

[3. Configuration example of information processing device]
Next, the configuration of the information processing apparatus 100 according to the embodiment will be described using FIG. FIG. 4 is a diagram illustrating a configuration example of the information processing apparatus 100 according to the embodiment. As shown in FIG. 4, the information processing apparatus 100 has a communication section 110, a storage section 120, and a control section .

(Communication unit 110)
The communication unit 110 is realized by, for example, a NIC (Network Interface Card) or the like. Also, the communication unit 110 is connected to the network N (see FIG. 3) by wire or wirelessly.

(storage unit 120)
The storage unit 120 is realized by, for example, a semiconductor memory device such as a RAM (Random Access Memory) or flash memory, or a storage device such as a hard disk or an optical disk. As shown in FIG. 4 , the storage unit 120 has a user database 121 , a sequence database 122 and a withdrawal period database 123 .

(User database 121)
The user database 121 stores various information about the user U. For example, the user database 121 stores various information such as user U attributes. FIG. 5 is a diagram showing an example of the user database 121. As shown in FIG. In the example shown in FIG. 5, the user database 121 includes items such as "user ID (Identifier)", "age", "gender", "home", "work place", "service", and "access log". have

“User ID” indicates identification information for identifying the user U. "Age" indicates the age of the user U identified by the user ID. Note that the “age” may be a specific age of the user U identified by the user ID, such as 35 years old. "Gender" indicates the gender of the user U identified by the user ID.

"Home" indicates location information of the home of the user U identified by the user ID. In the example shown in FIG. 5, "home" is represented by an abstract code such as "LC11", but may be latitude/longitude information or the like. Also, for example, "home" may be an area name or an address.

"Place of work" indicates location information of the place of work (school in the case of a student) of the user U identified by the user ID. In the example shown in FIG. 5, the "place of work" is illustrated as an abstract code such as "LC12", but may be latitude/longitude information or the like. Also, for example, the "place of work" may be an area name or an address.

"Service" indicates the service currently used by the user U identified by the user ID. Note that there may be multiple "services". That is, one user U may use multiple services at the same time.

“Access log” is an access log of the user U identified by the user ID, and indicates the access log of the user U acquired from each information providing device 200 .

For example, in the example shown in FIG. 5, the age of the user U identified by the user ID "U1" is "twenties" and the gender is "male". Also, for example, the user U identified by the user ID "U1" indicates that the home is "LC11". Also, for example, the user U identified by the user ID "U1" indicates that the place of work is "LC12". Also, for example, user U identified by user ID "U1" uses "service #11" and its access log is recorded in "log #1".

Here, in the example shown in FIG. 5, abstract values such as "U1", "LC11", "LC12", "service #11" and "log #1" are used, but "U1", " LC11", "LC12", "service #11" and "log #1" store information such as specific character strings and numerical values. Hereinafter, abstract values may also be illustrated in diagrams relating to other information.

The user database 121 is not limited to the above, and may store various types of information depending on the purpose. For example, the user database 121 may store various information about the terminal device 10 of the user U. FIG. In addition, the user database 121 stores attributes of the user U such as demographic (demographic attribute), psychographic (psychological attribute), geographic (geographical attribute), and behavioral (behavioral attribute). may store information about For example, the user database 121 includes name, family structure, occupation, position, income, qualification, residence type (detached house, condominium, etc.), presence or absence of car, commute/commuting time, commute/commuting route, commuter pass section (station , routes, etc.), stations with high frequency of use (other than the nearest station to home/work place), behavior history (location information with high frequency of use), lessons (place, time of day, etc.), hobbies, interests, lifestyle, etc. Information may be stored. The user database 121 may also store information on search queries (search keywords) that the user U has entered into a search engine or the like.

(Sequence database 122)
The sequence database 122 stores various information related to service usage transition sequence data created from an access log relating to service usage in the same session of the user U. FIG. FIG. 6 is a diagram showing an example of the sequence database 122. As shown in FIG. In the example shown in FIG. 6, the sequence database 122 has items such as "user ID", "sequence ID", and "service use transition sequence".

“User ID” indicates identification information for identifying the user U. "Sequence ID" indicates identification information for identifying service use transition sequence data. "Service use transition sequence" indicates each service included in the service use transition sequence data and the use order thereof. The information indicating each service may be identification information for identifying each service.

Furthermore, the 'service use transition sequence' includes items such as 'service' and 'use date and time'. "Service" indicates the service used by the user U. "Date and time of use" indicates the date and time when the user U used the service. The "date and time of use" is information indicating the order in which the user U used the service (for example, "1, 2, 3, ...", "A, B, C, ..."). There may be.

For example, in the example shown in FIG. 6, the service usage transition sequence data identified by the sequence ID "SQ1" is obtained from the access log regarding the usage of the service in the same session of the user U identified by the user ID "U1". Indicates that it has been created. Further, in the service usage transition sequence data identified by the sequence ID "SQ1", the services used by the user U are "first search site", "local summary site", "first map WEB", and "first search site". 1 map application”, “first search site”, “travel site”, and “accommodation reservation site”.

The sequence database 122 is not limited to the above, and may store various information depending on the purpose. For example, the sequence database 122 may store information regarding the date and time of use of each service included in the service use transition sequence data. In addition, the sequence database 122 may store information regarding the usage location of each service included in the service usage transition sequence data.

(Withdrawal period database 123)
The withdrawal period database 123 stores various information regarding the withdrawal period (absence time) of the target service. FIG. 7 is a diagram showing an example of the withdrawal period database 123. As shown in FIG. In the example shown in FIG. 7, the withdrawal period database 123 has items such as "sequence ID", "target service", and "withdrawal period".

“Sequence ID” indicates identification information for identifying service use transition sequence data. "Target service" indicates the target service included in the service use transition sequence data. Also, "withdrawal period" indicates the withdrawal period of the target service.

For example, in the example shown in FIG. 7, the withdrawal period of the target service "first map application" in the service use transition sequence data identified by the sequence ID "SQ1" is "withdrawal period #1".

It should be noted that the withdrawal period database 123 may store various information not limited to the above, depending on the purpose. For example, the withdrawal period database 123 may store information regarding the reason for withdrawal from the target service (the reason why the user stopped using the service). Also, the withdrawal period database 123 may store information on the squared error.

(control unit 130)
Returning to FIG. 4, the description is continued. The control unit 130 is a controller. Various programs (corresponding to an example of an information processing program) stored in an internal storage device are executed by using a storage area such as a RAM as a work area. In the example shown in FIG. 4 , the control unit 130 has a reception unit 131 , an acquisition unit 132 , a creation unit 133 , a learning unit 134 , an analysis unit 135 and a provision unit 136 .

(Reception unit 131)
The reception unit 131 receives registration of various information about each user U from the terminal device 10 of each user U via the communication unit 110 and stores the registration in the user database 121 .

(Acquisition unit 132)
The acquisition unit 132 acquires the access log of each user U from each information providing device 200 via the communication unit 110 .

(Creating unit 133)
The creation unit 133 creates service use transition sequence data regarding service use in the same session from the acquired access log of each user U. FIG.

Further, when the terminal device 10 of each user U is using a service provided by the information processing device 100, the creating unit 133 automatically creates an access log related to the service provided by the information processing device 100. may be created.

Further, the creation unit 133 may create service usage transition sequence data that includes usage of a predetermined target service and that the usage mode of another service satisfies a predetermined condition. For example, only sequences containing seasonal (depending on seasonal factors) services are extracted to create service usage transition sequence data for learning.

Note that the creation unit 133 may be integrated with the acquisition unit 132 . For example, the access log acquired by the acquisition unit 132 may be used as it is as the service usage transition sequence data. In addition, it can be considered that the creation unit 133 has acquired the service usage transition sequence data by creating the service usage transition sequence data.

(learning unit 134)
The learning unit 134 inputs the service use transition sequence data into the model, and makes the model learn about the user U's intention (purpose of use) of using the service in the same session. The learning unit 134 also creates a model for predicting the withdrawal period based on the learning result. At this time, the learning unit 134 causes the model to learn the withdrawal period as correct data. Although illustration is omitted, the learning unit 134 includes a calculation unit 134A and a generation unit 134B.

The calculator 134A calculates the squared error of the target service based on the service usage transition sequence data.

The generator 134B creates a model for predicting the withdrawal period of the target service based on the squared error of the target service. The generation unit 134B also generates withdrawal period information of the target service based on a model for predicting the withdrawal period.

The model is, for example, LSTM (Long short-term memory). However, LSTM is only an example. Any type of model can be adopted as the model.

Learning is, for example, deep learning using a deep neural network (DNN: Deep Neural Network). Data mining and other machine learning algorithms may also be used. The learning unit 134 learns the model using the various learning methods described above.

(Analysis unit 135)
The analysis unit 135 inputs all the learned service usage transition sequence data into the model and analyzes the reason for leaving the target service. Although illustration is omitted, the analysis unit 135 includes a rearrangement unit 135A and an estimation unit 135B.

The rearrangement unit 135A rearranges all the input service use transition sequence data in descending order of predicted withdrawal period. For example, the rearrangement unit 135A outputs the service usage transition sequence data in descending order of the withdrawal period. That is, the rearrangement unit 135A may be an output unit. At this time, the rearrangement unit 135A may limit the number of service usage transition sequence data to be output to a predetermined number. This reveals the characteristics of a transition sequence group with a long departure period (high probability of departure), that is, a transition sequence group including a cause of departure.

The estimation unit 135B estimates the cause of withdrawal included in the service usage transition sequence data. For example, the estimation unit 135B estimates a service that may cause the target service to leave from among the services included in the service usage transition sequence data.

(Providing unit 136)
The providing unit 136 provides information about the reason for leaving the target service to the outside via the communication unit 110 . For example, the providing unit 136 may provide a report on the reason for leaving the target service in response to an external request. Note that the providing unit 137 may function as a display control unit that causes a display device such as a display to display information about the reason for withdrawal from the target service.

Further, the providing unit 136 may provide the information on the service use transition sequence group with high probability of leaving to the outside via the communication unit 110 . For example, the providing unit 136 may provide information about a service use transition sequence group with a high probability of leaving to the outside in response to a request from the outside.

Also, the providing unit 136 may provide various services to the terminal device 10 of each user U via the communication unit 110 .

[4. Processing procedure]
Next, a processing procedure performed by the information processing apparatus 100 according to the embodiment will be described with reference to FIG. FIG. 8 is a flow chart showing a processing procedure according to the embodiment. Note that the processing procedure described below is repeatedly executed by the control unit 130 of the information processing apparatus 100 .

The control unit 130 of the information processing device 100 acquires the access log of each user U from each information providing device 200 via the communication unit 110 (step S101).

The control unit 130 creates service use transition sequence data regarding service use in the same session from the access log of each user U (step S102).

The control unit 130 inputs the service use transition sequence data into the model, and makes the model learn the intention (usage purpose) of the user U using the service in the same session (step S103). In practice, a model for predicting the withdrawal period may be learned from the beginning without learning the purpose of use.

The control unit 130 inputs the service usage transition sequence data to the model, and calculates the squared error, which is the error between the actual value and the predicted value by the model (step S104).

The control unit 130 performs model learning so as to reduce the squared error, and creates a model for predicting the withdrawal period (step S105).

The control unit 130 sets a certain percentage of the learning data in advance as test data, inputs all the service usage transition sequence data of the test data into the model, and sorts the predicted withdrawal periods in descending order. Rearrange (step S106). This reveals the characteristics of a transition sequence group with a long departure period (high probability of departure), that is, a transition sequence group including a cause of departure.

The control unit 130 estimates the cause of withdrawal from the target service from the transition sequence group with a long withdrawal period (high probability of withdrawal) (step S107).

[5. Modification]
The terminal device 10 and the information processing device 100 described above may be implemented in various different forms other than the above embodiments. So, below, the modification of embodiment is demonstrated.

In the above-described embodiment, service usage transition sequence data may be created from an access log relating to service usage in the same session of a user U having the same attribute. For example, since the service usage transition sequence is considered to differ depending on the attributes of the user U (gender/age group/residential area/family composition/hobbies, etc.), service usage transition sequence data classified according to the attributes of the user U may be created. As a result, a model is created for each user U attribute.

Further, in the above-described embodiment, service usage transition sequence data may be created from an access log regarding service usage in the same session according to the situation (context) when the user U uses the service. For example, the reasons for withdrawal from the service may change depending on the situation and environment in which user U is placed (move, purchase a car, advance to higher education/employment, change jobs, get married, give birth/childcare, etc.). Service usage transition sequence data classified according to the situation may be created. As a result, a model is created for each situation in which the user U is placed.

In addition, in the above embodiment, a model for predicting the "withdrawal period" is created. A model for estimating the "impression of the user when using the service in the order indicated by the sequence data" may be learned.

For example, it can be estimated that the satisfaction level is high when a map service is used from a web search, but the impression is poor when a map service is used from shopping.

Also, for example, the information processing apparatus 100 uses a technique such as crowdsourcing to acquire sequence data and a value indicating an impression when using the target service in the sequence from the user. Here, the value indicating the impression is, for example, a value ranging from 0 points (unsatisfactory) to 10 points (satisfactory). Then, the information processing apparatus 100 performs model pre-training and satisfaction level learning using the techniques described above.

A model that has undergone such learning processing can estimate, from the sequence data, the user's degree of satisfaction with the target service in the sequence indicated by the sequence data.

In addition to the various evaluations input by the user, the value indicating the impression is, for example, the user's expression, pulse, etc. when using the target service in the sequence indicated by the sequence data using a camera, etc. It may be a value estimated from facial expressions or biometric information obtained by acquiring various biometric information.

Further, the information processing apparatus 100 may use the withdrawal period as information suggesting the user's impression. That is, the information processing apparatus 100 calculates a score indicating the user's impression using various known techniques for estimating the user's impression, and performs model learning so as to estimate the calculated score. good too.

In the above embodiment, the order is learned by inputting the information indicating the services in sequence, but the invention is not limited to this. For example, a non-recursive model such as a CNN may be learned by inputting information indicating a service and numerical values indicating an order into a model and outputting a withdrawal period and user's impression.

[6. effect〕
As described above, the information processing apparatus 100 according to the present application has an acquisition unit and a learning unit. The acquiring unit acquires sequence data indicating transition of service usage of each user and including usage of a predetermined target service. The learning unit makes the model learn from the sequence data to output withdrawal period information indicating a period from when the user uses the target service until when the user uses the target service again. This makes it possible to find out the cause of withdrawal more effectively.

The learning unit trains a model having an LSTM (Long short-term memory) structure. This makes it possible to estimate the reason for leaving the analysis target service based on the sequence data, which is time-series data.

The acquisition unit acquires, as sequence data, data indicating a plurality of services used by the user and an order in which the services were used. The learning unit causes the model to learn to output withdrawal period information when information indicating each service used by the user is input to the model in the order in which each service was used. As a result, it is possible to carry out specific learning for finding out the cause of withdrawal more effectively.

The learning unit causes the model to learn such that when information indicating the service used by the user is input, the model outputs information indicating the service to be used next by the user, and then, in the order in which each service is used. The model is trained to output withdrawal period information when inputting to the model. This allows pre-training of the model for the intended use of the target service.

The learning unit learns the model so as to output, as the withdrawal period information, the number of other services used by the user from the time the user uses the target service until the user uses the target service again. As a result, the number of contents used during the withdrawal period can be used instead of the time.

The learning unit learns the model so as to output the elapsed time from when the user uses the target service until when the user uses the target service again as the withdrawal period information. As a result, the time elapsed during the period can be used as the withdrawal period.

The acquisition unit acquires sequence data that includes use of a predetermined target service and that the usage mode of another service satisfies a predetermined condition. This allows for context filtering. For example, sequence data for learning can be created by extracting only sequences that include seasonal services.

Moreover, the information processing apparatus 100 according to the present application further includes a generation unit and an output unit. The generation unit generates separation period information corresponding to each piece of sequence data using the model trained by the learning unit. The output unit outputs sequence data whose period indicated by the separation period information satisfies a predetermined condition. Thereby, the sequence data can be output according to the detachment period.

The output unit outputs the sequence data in descending order of the period indicated by the separation period information. Thereby, the sequence data can be rearranged in descending order of the detachment period.

The output unit outputs a predetermined number of pieces of sequence data in descending order of the period indicated by the separation period information. This makes it possible to limit the number of output sequence data. As a result, only the sequence data with the longer detachment period can be targeted.

The acquisition unit acquires, as sequence data, sequence data using each function of one application as a service. This makes it possible to target fine-grained sequence data. As a result, it is possible to estimate the cause of withdrawal from the function of one application.

From another point of view, the information processing apparatus 100 according to the present application has an acquisition unit and a learning unit. The obtaining unit obtains sequence data indicating transition of use of each user's service and including use of a predetermined target service. The learning unit causes the model to learn from the sequence data so as to output index information that is an index of the user's impression of the target service. This makes it possible to estimate the user's satisfaction/dissatisfaction.

[7. Hardware configuration]
Also, the terminal device 10 and the information processing device 100 according to the above-described embodiments are implemented by a computer 1000 configured as shown in FIG. 9, for example. The information processing apparatus 100 will be described below as an example. FIG. 9 is a diagram illustrating an example of a hardware configuration; The computer 1000 is connected to an output device 1010 and an input device 1020, and an arithmetic device 1030, a primary storage device 1040, a secondary storage device 1050, an output I/F (Interface) 1060, an input I/F 1070, and a network I/F 1080 are buses. It has a form connected by 1090.

The arithmetic device 1030 operates based on programs stored in the primary storage device 1040 and the secondary storage device 1050, programs read from the input device 1020, and the like, and executes various processes. The arithmetic unit 1030 is implemented by, for example, a CPU (Central Processing Unit), MPU (Micro Processing Unit), ASIC (Application Specific Integrated Circuit), FPGA (Field Programmable Gate Array), or the like.

The primary storage device 1040 is a memory device such as a RAM (Random Access Memory) that temporarily stores data used for various calculations by the arithmetic device 1030 . The secondary storage device 1050 is a storage device in which data used for various calculations by the arithmetic device 1030 and various databases are registered. State Drive), flash memory, or the like. The secondary storage device 1050 may be an internal storage or an external storage. Also, the secondary storage device 1050 may be a removable storage medium such as a USB memory or an SD (Secure Digital) memory card. Also, the secondary storage device 1050 may be a cloud storage (online storage), a NAS (Network Attached Storage), a file server, or the like.

The output I/F 1060 is an interface for transmitting information to be output to the output device 1010 that outputs various information such as a display, a projector, and a printer. (Digital Visual Interface), HDMI (registered trademark) (High Definition Multimedia Interface), and other standardized connectors. Also, the input I/F 1070 is an interface for receiving information from various input devices 1020 such as a mouse, keyboard, keypad, buttons, scanner, etc., and is realized by, for example, USB.

Also, the output I/F 1060 and the input I/F 1070 may be wirelessly connected to the output device 1010 and the input device 1020, respectively. That is, output device 1010 and input device 1020 may be wireless devices.

Also, the output device 1010 and the input device 1020 may be integrated like a touch panel. In this case, the output I/F 1060 and the input I/F 1070 may also be integrated as an input/output I/F.

Note that the input device 1020 includes, for example, optical recording media such as CDs (Compact Discs), DVDs (Digital Versatile Discs), PDs (Phase change rewritable discs), magneto-optical recording media such as MOs (Magneto-Optical discs), and tapes. It may be a device that reads information from a medium, a magnetic recording medium, a semiconductor memory, or the like.

Network I/F 1080 receives data from other devices via network N and sends the data to arithmetic device 1030, and also transmits data generated by arithmetic device 1030 via network N to other devices.

Arithmetic device 1030 controls output device 1010 and input device 1020 via output I/F 1060 and input I/F 1070 . For example, arithmetic device 1030 loads a program from input device 1020 or secondary storage device 1050 onto primary storage device 1040 and executes the loaded program.

For example, when the computer 1000 functions as the information processing device 100 , the arithmetic device 1030 of the computer 1000 implements the functions of the control unit 130 by executing a program loaded on the primary storage device 1040 . Further, arithmetic device 1030 of computer 1000 may load a program acquired from another device via network I/F 1080 onto primary storage device 1040 and execute the loaded program. Further, the arithmetic unit 1030 of the computer 1000 may cooperate with another device via the network I/F 1080, and call functions, data, etc. of the program from another program of the other device for use.

[8. others〕
Although the embodiments of the present application have been described above, the present invention is not limited by the contents of these embodiments. In addition, the components described above include those that can be easily assumed by those skilled in the art, those that are substantially the same, and those within the so-called equivalent range. Furthermore, the components described above can be combined as appropriate. Furthermore, various omissions, replacements, or modifications of components can be made without departing from the gist of the above-described embodiments.

Further, among the processes described in the above embodiments, all or part of the processes described as being automatically performed can be manually performed, or the processes described as being performed manually can be performed manually. All or part of this can also be done automatically by known methods. In addition, information including processing procedures, specific names, various data and parameters shown in the above documents and drawings can be arbitrarily changed unless otherwise specified. For example, the various information shown in each drawing is not limited to the illustrated information.

Also, each component of each device illustrated is functionally conceptual, and does not necessarily need to be physically configured as illustrated. In other words, the specific form of distribution and integration of each device is not limited to the illustrated one, and all or part of them can be functionally or physically distributed and integrated in arbitrary units according to various loads and usage conditions. Can be integrated and configured.

For example, the information processing apparatus 100 described above may be implemented by a plurality of server computers, and depending on the function, may be implemented by calling an external platform or the like using an API (Application Programming Interface), network computing, or the like. can be changed flexibly.

Also, the above-described embodiments and modifications can be appropriately combined within a range that does not contradict the processing content.

Also, the above-mentioned "section, module, unit" can be read as "means" or "circuit". For example, the acquisition unit can be read as acquisition means or an acquisition circuit.

1 information processing system 10 terminal device 100 information processing device 110 communication unit 120 storage unit 121 user database 122 sequence database 123 withdrawal period database 130 control unit 131 reception unit 132 acquisition unit 133 creation unit 134 learning unit 135 analysis unit 136 provision unit 200 Information provision device

Claims (14)

an acquisition unit for acquiring sequence data indicating the transition of service usage by each user, the sequence data including usage of a predetermined target service;
a learning unit that causes a model to learn so as to output, from the sequence data, withdrawal period information indicating a period from when the user uses the target service until when the target service is used again. information processing equipment. The information processing apparatus according to claim 1, wherein the learning unit learns a model having an LSTM configuration as the model. The acquisition unit acquires, as the sequence data, data indicating a plurality of services used by the user and an order in which the services were used,
The learning unit causes the model to learn such that when information indicating each service used by the user is input to the model in the order in which each service was used, the model outputs the withdrawal period information. 3. The information processing apparatus according to claim 1 or 2. The learning unit causes the model to learn such that when information indicating a service used by the user is input, the model outputs information indicating a service to be used next by the user, and then uses each service. 4. The information processing apparatus according to claim 3, wherein the model is made to learn so as to output the withdrawal period information when the information is input to the model in the order in which the information is input to the model. The learning unit causes the model to learn so as to output, as the withdrawal period information, the number of other services used by the user from the time the user uses the target service until the user uses the target service again. The information processing apparatus according to any one of claims 1 to 4, characterized in that: The learning unit causes the model to learn so as to output, as the withdrawal period information, the time elapsed from when the user used the target service until when the user used the target service again. The information processing apparatus according to any one of claims 1 to 4. 7. The acquisition unit acquires sequence data that includes use of the predetermined target service and that satisfies a predetermined condition in a usage mode of another service. The information processing device according to . a generation unit that generates the separation period information corresponding to each sequence data using the model trained by the learning unit;
8. The information processing apparatus according to any one of claims 1 to 7, further comprising: an output unit that outputs sequence data that satisfies a predetermined condition for a period indicated by said withdrawal period information. 9. The information processing apparatus according to claim 8, wherein the output unit outputs the sequence data in descending order of the period indicated by the separation period information. 10. The information processing apparatus according to claim 9, wherein the output unit outputs a predetermined number of the sequence data in descending order of the period indicated by the separation period information. The information processing apparatus according to any one of claims 1 to 10, wherein the acquisition unit acquires, as the sequence data, sequence data using each function of one application as a service. An information processing method executed by an information processing device,
an obtaining step of obtaining sequence data indicating the transition of use of each user's service, the sequence data including use of a predetermined target service;
and a learning step of causing a model to learn from the sequence data to output withdrawal period information indicating a period from when the user uses the target service until when the user uses the target service again. information processing method. an acquisition procedure for acquiring sequence data indicating the transition of use of each user's service, the sequence data including use of a predetermined target service;
and a learning procedure for causing a computer to learn from the sequence data to output withdrawal period information indicating the period from when the user uses the target service until when the user uses the target service again. information processing program. an acquisition unit for acquiring sequence data indicating the transition of service usage by each user, the sequence data including usage of a predetermined target service;
and a learning unit that causes a model to learn so as to output index information, which is an index of the user's impression of the target service, from the sequence data.

Images