JP7620906B2 - Behavioral pattern analysis device, method, and program - Google Patents

Description

The present invention relates to a behavioral pattern analysis device, method, and program for analyzing a user's behavioral patterns regarding viewable information viewable on the web.

In recent years, it has become common to provide users with information about products and services in viewable information that is viewable on the web, such as web pages on websites and viewable information displayed on applications on smartphones and the like, and to encourage users to take action such as purchasing the products or services. In order to encourage users to take action in this way, a technique is known that analyzes the history of users' behavior regarding viewable information.

For example, Patent Literature 1 proposes an information analysis device that includes a log acquisition means for acquiring an action log when a user identified by a user ID accesses a service providing site using a user terminal, a clustering means for classifying action patterns based on the action log acquired by the log acquisition means into at least one cluster, and a machine learning means for performing machine learning using multiple features including the cluster information classified by the clustering means to generate a discriminant function (Patent Literature 1).

JP 2016-062411 A

However, the information analysis device of Patent Document 1 first uses the cluster information classified by clustering to perform machine learning to generate a discriminant function. Therefore, if there is a bias in the classified clusters, even if machine learning is then performed to generate a discriminant function, the discriminant function will be biased and the analysis results will also be biased or incomplete, which may prevent an appropriate analysis.

In view of these problems, the present invention aims to provide a behavioral pattern analysis device, method, and program that suppresses bias in user clustering and enables appropriate analysis.

The present invention provides the following solutions.
The present invention provides a behavioral pattern analysis device for analyzing a user's behavioral pattern with respect to viewable information viewable on the web, comprising:
a history information acquiring means for acquiring, for each user, history information indicating actions regarding access to the viewable information by each individual user, together with user information regarding the user;
A specific user extraction means for extracting a specific user who is a user having the history information indicating a certain behavior;
A clustering means for clustering the specific user into a plurality of clusters using feature amounts of a plurality of types of behavioral patterns by unsupervised learning of machine learning;
A clustering specific user extraction means for extracting the clustered specific users for each cluster;
and a display control means for displaying, for each cluster, the user information of the specific users clustered on a display means.

The present invention may further include a clustering behavior pattern identification means for identifying tendencies in the behavior patterns of the clustered specific users for each cluster.

In addition, in the present invention, the display control means may display on the display means specific user count information regarding the number of all the specific users extracted by the specific user extraction means, and clustering specific user count information regarding the number of the specific users extracted by the clustering specific user extraction means.

In addition, although the present invention is in the category of an apparatus, it may also be a method and a program.

According to the present invention, it is possible to reduce bias in user clustering and perform appropriate analysis.

FIG. 1 shows an example of a display of the results of clustering performed by a behavior pattern analysis device according to this embodiment. FIG. 2 is a block diagram showing the functional configuration of the behavior pattern analysis device of this embodiment. FIG. 3 is a diagram illustrating the user information DB. FIG. 4 is a diagram illustrating the history information DB. FIG. 5 is a diagram for explaining an outline of clustering by the clustering means of the behavior pattern analysis device of this embodiment. FIG. 6 is a flowchart showing an example of the behavior pattern analysis process of this embodiment. FIG. 7 shows an example of a specific user extraction screen displayed by the behavior pattern analysis device according to this embodiment. FIG. 8 shows an example of a display of a clustering specific user extraction screen by the behavior pattern analysis device according to this embodiment.

The best mode for carrying out the present invention will be described below with reference to the drawings. Note that this is merely an example, and the technical scope of the present invention is not limited to this example.

<Overall composition>
FIG. 1 shows an example of a display of the results of clustering performed by a behavior pattern analysis device according to this embodiment.
The behavior pattern analysis device of this embodiment analyzes user behavior patterns regarding viewable information that can be viewed on the web (e.g., websites that can be viewed using the Internet, displays on applications that sell products, etc. using the Internet, etc.).

Specifically, the behavior pattern analysis device acquires, for each user, historical information indicating the behavior of each individual user regarding access to the viewable information, together with user information regarding the user.

The behavioral pattern analysis device then extracts from all users specific users who have historical information indicating the behavior (e.g., conversion (such as purchasing a product)) desired by an operator (e.g., a person selling a product, a person analyzing user behavior, or a person encouraging users to take a desired action).

Then, the behavioral pattern analysis device uses unsupervised machine learning to cluster the extracted specific users into multiple clusters using features of multiple types of behavioral patterns (data of each specific user (CV number, CV frequency, PV number, PV frequency, etc.)).

The behavioral pattern analysis device calculates the clusters resulting from the clustering (in the example shown in Figure 1, "Not Counted," "Pattern A," "Pattern B," "Pattern C," "Pattern D," "Pattern E," and "Pattern F") and the number of specific users classified into each cluster, and displays them as shown in Figure 1.

In addition, when a cluster is selected from the list of clusters shown in FIG. 1 based on a user's operation, for example, the behavior pattern analysis device displays user information of a specific user who is clustered in that cluster.

In this way, the behavior pattern analysis device extracts specific users who have historical information including a certain behavior pattern, and by unsupervised machine learning, clusters the specific users into multiple clusters using features of multiple types of behavior patterns, extracts the clustered specific users for each cluster, and displays the user information of the clustered specific users for each cluster on the display means.

Here, if a user is labeled in advance with a specific behavioral pattern (e.g., accessing during a specific time period) and the association between the labeled user's behavioral pattern and the behavior desired by the operator is extracted using supervised learning in machine learning, it will not be possible to analyze the association between a user who does not have a specific behavioral pattern in advance and the behavior desired by the operator, which could result in bias or omissions in the analysis results.

According to the behavioral pattern analysis device of this embodiment, specific users who have behaved in a certain manner are clustered by unsupervised machine learning. For example, specific users who have behaved in a manner desired by an operator are first extracted, and then the clustering is performed by unsupervised machine learning, thereby making it possible to reduce bias in the analysis results.
Therefore, bias in user clustering can be reduced, enabling appropriate analysis.

<Functional configuration>
FIG. 2 is a block diagram showing the functional configuration of the behavior pattern analysis device of this embodiment.
The behavior pattern analysis device 1 is, for example, configured as a server, and includes a processor, memory, and storage (not shown), which are connected by a bus. The processor is, for example, configured as a CPU (Central Processing Unit), and functions as each of the means described below by reading and executing various programs stored in the memory and performing various processes. The memory stores the programs executed by the CPU, and is, for example, configured as a ROM (Read Only Memory) and a RAM (Random Access Memory). The storage stores databases, control programs, and the like, which will be described later.
The behavior pattern analysis device 1 does not need to be configured as a single server, and various processes may be executed in whole or in part by, for example, a terminal (e.g., a personal computer, a mobile terminal, etc.) operated by an administrator of the behavior pattern analysis device 1, or a server connected to such a terminal via a network.

The behavior pattern analysis device 1 includes a user information acquisition means 11, a user information database (hereinafter also referred to as "user information DB") 12, a history information acquisition means 13, a history information database (hereinafter also referred to as "history information DB") 14, a specific user extraction means 15, a clustering means 16, a clustering specific user extraction means 17, a display means 18, and a display control means 19.

The user information acquisition means 11 acquires user information from a terminal operated by the user (e.g., a personal computer, a mobile terminal, etc.) or an external device such as an offline database (hereinafter also referred to as an "offline DB") 200, and stores the user information in the user information DB 12.

FIG. 3 is a diagram illustrating the user information DB.
The user information DB 12 stores user information in which identification information for identifying each user is associated with age, location (the prefecture where the user lives, address, etc.), etc. In addition, the user information may also include gender, user name (e.g., a nickname set by the user, the user's name, etc.), user attributes (e.g., membership rank of a group operated by a website operator, number of PV (Page Views), number of CV (Conversions), etc.), offline data (e.g., actual behavior history of the user, NPS (Net Promoter Score), business index, etc.), etc. In this case, these pieces of information may be acquired by the user information acquisition means 11 and stored in the user information DB 12 in association with the identification information.

Returning to FIG. 2, the history information acquisition means 13 acquires, for each user, history information including the viewable information accessed by the user, the visit date and time indicating the time when the viewable information was accessed, information indicating the terminal used by the user to access the viewable information (e.g., a personal computer (PC) or a mobile terminal (smartphone)), information indicating the event type indicating the user's behavior in the viewable information (e.g., whether it was only PV or led to CV), the stay time indicating the time when the viewable information was accessed, details of the event type (e.g., the content of the CV, the introduction destination of the service or product), etc., from a device (server, etc.) accessed from the user's terminal, an offline DB 200 connected to the behavior pattern analysis device 1 via a network, or an external database (hereinafter also referred to as "external DB") 300 connected to the behavior pattern analysis device 1 via a network, and stores the information in the history information DB 14. In addition, the history information acquisition means 13 generates user count information regarding the number of all users who have acquired user information.

FIG. 4 is a diagram illustrating the history information DB.
The history information DB 14 stores, for each user, the visit date and time indicating the time when the viewable information was accessed by the user, information indicating the terminal used by the user to access the viewable information, information indicating the event type indicating the user's behavior in the viewable information, and the stay time indicating the time when the viewable information was accessed, in association with the viewable information accessed by the user. As shown in Fig. 4, the history information DB 14 stores history information for each visit date and time to certain viewable information. In other words, the history information acquisition means 13 also stores, for each user, in the history information DB 14, the number of times certain viewable information has been accessed and the frequency of how many times certain viewable information has been accessed in a certain period of time.

In the example shown in Figure 4, the name of a web page ("Ug special page" in the example shown in Figure 4) is shown as an example of "viewable information", but "viewable information" can be any information that allows a user to identify viewable information using the Internet, such as the address of a web page, the name of a shop on a website, or the name of an application that sells products, etc., using the Internet.

The history information may also include, as information regarding user operations on each piece of viewable information, scroll information indicating the amount scrolled on the display screen and the scrolling speed, and pointer information indicating the path, amount of movement, and speed of a pointer that can be moved on the display screen by mouse operations or touch operations on the terminal operated by the user. In this case, these pieces of information may be acquired by the history information acquisition means 13 and stored in association with the viewable information in the history information DB 14. The history information acquisition means 13 may also classify and acquire information indicating the inflow path into the user's first access and the second or subsequent access, and store each of them, or may classify and acquire information indicating the inflow path when the CV was reached into the case of only PV, and store each of them.

Returning to FIG. 2, the specific user extraction means 15 extracts specific users who have history information indicating a certain behavior. In detail, when the operator specifies a desired behavior, the specific user extraction means 15 refers to the history information DB 14 (see FIG. 4) and extracts users who have history information indicating the specified behavior (e.g., CV: purchase, etc.) as specific users. In addition, the specific user extraction means 15 generates specific user count information regarding the number of all extracted specific users.

The clustering means 16 uses unsupervised machine learning to cluster the specific users extracted by the specific user extraction means 15 into multiple clusters using features of multiple types of behavioral patterns based on the history information stored in the history information DB 14 (see Figure 4).

FIG. 5 is a diagram for explaining an outline of clustering by the clustering means of the behavior pattern analysis device of this embodiment.
The clustering means 16 refers to the user information DB 12 (see Figure 3) and the history information DB 14 (see Figure 4), and clusters the history information of the specific user extracted by the specific user extraction means 15 into multiple clusters using unsupervised learning of known machine learning.
For example, the clustering means 16 distributes the specific users based on the historical information of the specific users by the features of their behavioral patterns (data of each specific user (CV number, CV frequency, PV number, PV frequency, etc.) and groups them into a predetermined number of clusters (six in the example shown in FIG. 5 ), and finds the center of distribution in each group. Then, the clustering means 16 regroups the specific users based on their proximity to the found center, and repeats the process of finding the center of the regroups to form clusters with the predetermined number of clusters. Note that, although the number of clusters is six in this embodiment, this is not limited to this, and the number of clusters may be determined arbitrarily.

Returning to FIG. 2, the clustering specific user extraction means 17 extracts specific users clustered by the clustering means 16. In detail, when the operator specifies a desired cluster, the clustering specific user extraction means 17 extracts specific users clustered to the specified cluster by the clustering means 16. The clustering specific user extraction means 17 also counts the number of specific users for each extracted cluster, and generates clustering specific user number information indicating the number of specific users counted and the ratio of the number of specific users for each cluster to the number of all specific users extracted by the specific user extraction means 15. The clustering specific user extraction means 17 also counts the number of specific users not clustered by the clustering means 16 (e.g., specific users whose history information is acquired by the history information acquisition means 13 after clustering by the clustering means 16), and generates uncounted specific user number information indicating the number of uncounted specific users counted and the ratio of the number of uncounted specific users to the number of all specific users extracted by the specific user extraction means 15.

The display means 18 is composed of, for example, a display or a touch panel, and displays a screen showing the clustering results (see Figure 1) under the control of the display control means 19.

The display control means 19 displays, on the display means 18, the user information of the specific users clustered for each cluster by the clustering means 16 based on the operation of the operator. The display control means 19 also displays, on the display means 18, the specific user count information generated by the specific user extraction means 15, the clustered specific user count information generated by the clustering specific user extraction means 17, and the uncounted specific user count information.

<Behavioral pattern analysis processing>
Next, an example of a behavior pattern analysis process performed by the behavior pattern analysis device will be described with reference to FIG.
FIG. 6 is a flowchart showing an example of the behavior pattern analysis process of this embodiment.
In step S<b>1 , the user information acquisition unit 11 acquires user information from a terminal operated by a user or from the offline DB 200 or the like which is an external device, and stores the user information in the user information DB 12 .

In step S2, the history information acquisition means 13 acquires, for each user, history information including the viewable information accessed by the user, the visit date and time indicating the time when the viewable information was accessed, information indicating the terminal used by the user to access the viewable information, information indicating the event type indicating the user's behavior in the viewable information, and the stay time indicating the time when the viewable information was accessed, from the device accessed from the user's terminal, the offline DB 200, or the external DB 300 by displaying the viewable information on the user's terminal, and stores the information in the history information DB 14.

In step S3, the specific user extraction means 15 extracts specific users who have history information indicating a certain behavior.

FIG. 7 shows an example of a specific user extraction screen displayed by the behavior pattern analysis device according to this embodiment.
In the behavior pattern analysis device 1, user information is associated with history information, and for example, when an operator selects a certain action (e.g., a conversion (e.g., "apply for seminar" or "inquire about Ug") on an initial screen arranged in chronological order, the specific user extraction means 15 extracts a specific user who is a user having history information indicating the certain action (e.g., a conversion). Then, the display control means 19 displays on the display means 18 a specific user extraction screen as shown in FIG. 7, in which the user information of the specific user is associated with history information related to the action selected by the operator. The display control means 19 also displays the specific user count information generated by the specific user extraction means 15 on the specific user extraction screen ("208 people" in the example shown in FIG. 7).

Returning to FIG. 6, in step S4, the specific users extracted by the specific user extraction means 15 are clustered into multiple clusters using the features of multiple types of behavioral patterns through unsupervised machine learning.

In the examples shown in Fig. 1 and Fig. 7, the operator can click the "Select behavior pattern" button on the specific user extraction screen shown in Fig. 7 to display a pull-down menu as shown in Fig. 1, and in this pull-down menu, the operator can select one of the clusters obtained by clustering: "Not counted", "Pattern A", "Pattern B", "Pattern C", "Pattern D", "Pattern E", or "Pattern F". At this time, the number of specific users belonging to each cluster (clustered specific user number information) is displayed in association with each cluster.

Then, when the operator selects one of the multiple clusters ("Not Counted", "Pattern A", "Pattern B", "Pattern C", "Pattern D", "Pattern E", or "Pattern F") in a pull-down menu such as that shown in FIG. 1, the behavior pattern analysis device 1 proceeds to step S5.

Returning to FIG. 6, in step S5, the clustering specific user extraction means 17 extracts specific users of the cluster that was clustered by the clustering means 16 in step S4 and selected by the operator.

FIG. 8 shows an example of a display of a clustering specific user extraction screen by the behavior pattern analysis device according to this embodiment.
As shown in Fig. 8, when the operator selects any one of a plurality of clusters ("Not Counted", "Pattern A", "Pattern B", "Pattern C", "Pattern D", "Pattern E", "Pattern F") in the pull-down menu as shown in Fig. 1, the display control means 19 displays, in step S5, a clustering specific user extraction screen on the display means 18, which lists the specific users of the cluster ("Pattern A" in the example shown in Fig. 8) that has been clustered by the clustering means 16 and selected by the operator. The display control means 19 also displays the clustering specific user number information generated by the clustering specific user extraction means 17 ("133 people (63.9%)" in the example shown in Fig. 8) on the clustering specific user extraction screen.

＜Effects＞
According to the behavior pattern analysis device of the present embodiment described above, specific users who have historical information including a certain behavior pattern are extracted, and the specific users are clustered into a plurality of clusters using features of a plurality of types of behavior patterns by unsupervised machine learning. The clustered specific users are extracted for each cluster, and the user information of the clustered specific users for each cluster is displayed on a display means.
This makes it possible to cluster specific users who have behaved in a certain way using unsupervised machine learning, for example, by first extracting specific users who have behaved in a way that an operator desires, and then clustering them using unsupervised machine learning, without incorporating the operator's subjectivity.
Therefore, bias in user clustering can be reduced, enabling appropriate analysis.

Also displayed are specific user number information relating to the number of all specific users extracted by the specific user extraction means, and clustering specific user number information relating to the number of specific users extracted by the clustering specific user extraction means.
This allows the operator to compare the number of users belonging to a certain cluster with the total number of users who performed a certain behavior, making it possible to prioritize multiple clusters in subsequent analysis, resulting in a more appropriate analysis.

Although the embodiments of the present invention have been described above, the present invention is not limited to these embodiments. Furthermore, the effects described in the embodiments of the present invention are merely a list of the most favorable effects resulting from the present invention, and the effects of the present invention are not limited to those described in the embodiments of the present invention.

1... Behavioral pattern analysis device, 11... User information acquisition means, 12... User information DB, 13... History information acquisition means, 14... History information DB, 15... Specific user extraction means, 16... Clustering means, 17... Clustering specific user extraction means, 18... Display means, 19... Display control means, 200... Offline DB, 300... External DB

Claims (4)

A behavioral pattern analysis device that analyzes a user's behavioral pattern with respect to viewable information viewable on the web, comprising:
a history information acquiring means for acquiring, for each user, history information indicating actions regarding access to the viewable information by each individual user, together with user information regarding the user;
A specific user extraction means for extracting a specific user who is a user having the history information indicating a certain behavior;
A clustering means for clustering the specific user into a plurality of clusters using feature amounts of a plurality of types of behavioral patterns by unsupervised learning of machine learning;
A clustering specific user extraction means for extracting the specific user clustered into the cluster designated by an operator from among the plurality of clusters;
A behavior pattern analysis device comprising: a display control means for displaying, on a display means, a clustering specific user extraction screen that arranges the user information and history information of the specific users clustered into the cluster specified by an operator . The behavioral pattern analysis device according to claim 1, wherein the display control means displays on the display means specific user number information regarding the number of all the specific users extracted by the specific user extraction means and clustering specific user number information regarding the number of the specific users extracted by the clustering specific user extraction means. A method executed by a behavior pattern analysis device that analyzes a user's behavior pattern with respect to viewable information viewable on the web, comprising:
acquiring, for each user, history information indicating an action of each user with respect to access to the viewable information, together with user information regarding the user;
Extracting a specific user who has the history information indicating a certain behavior;
A step of clustering the specific user into a plurality of clusters using feature amounts of a plurality of types of behavioral patterns by unsupervised learning of machine learning;
extracting the specific user clustered into the cluster designated by an operator from among the plurality of clusters;
and displaying on a display means a clustering specific user extraction screen which arranges the user information and history information of the specific users clustered into the cluster specified by an operator . A behavioral pattern analysis device for analyzing a user's behavioral pattern with respect to viewable information viewable on the web,
a history information acquiring means for acquiring, for each user, history information indicating actions regarding access to the viewable information by each individual user, together with user information regarding the user;
A specific user extraction means for extracting a specific user who is a user having the history information indicating a certain behavior;
A clustering means for clustering the specific users into a plurality of clusters using feature amounts of a plurality of types of behavioral patterns by unsupervised learning of machine learning;
a clustering specific user extraction means for extracting the specific user clustered into the cluster designated by an operator from among the plurality of clusters;
A program that functions as a display control means for displaying, on a display means, a clustering specific user extraction screen that arranges the user information and history information of the specific users clustered into the cluster specified by an operator .