JP6498345B1 - Information processing apparatus, program, and display control method - Google Patents

Abstract

To provide a technique capable of increasing a conversion rate by appropriately guiding a user's page transition in a website.
A storage unit that stores a learned model that outputs value information indicating the value of transition to a web page for a web page that can be transitioned in the website, and a predetermined web in the website By inputting, into the learned model, a first acquisition unit that acquires state information indicating a state of a user who refers to a page, information indicating a predetermined web page, and state information acquired by the first acquisition unit, A second acquisition unit that acquires value information about each of a plurality of transition destination web pages that can transition from a predetermined web page, and a display control unit that controls display of the website, acquired by the second acquisition unit A display control unit that displays information for guiding the user to one of the plurality of transition destination web pages on a predetermined web page based on the value information.
[Selection] Figure 2

Description

  The present invention relates to an information processing apparatus, a program, and a display control method.

  The conversion rate is known as an indicator for websites. The conversion rate is an index indicating the ratio of users who have reached a predetermined target such as a commercial transaction among users who have visited a website. Websites have been devised in various ways to increase the conversion rate. Patent Document 1 discloses a technique that can increase the conversion rate by optimizing the arrangement of the advertising space in the content.

Japanese Patent No. 6346684

  For example, a user who visits a website to conduct a commercial transaction collects information on products and services by referring to various pages in the website, and finally determines whether or not to conduct a commercial transaction. However, even when a user visits a website, it is considered that there are many cases where the user does not arrive at information useful for the determination of the commercial transaction and leaves the website before the commercial transaction.

  Then, an object of this invention is to provide the technique which can raise a conversion rate by guide | inducing the user's page transition in a website appropriately.

  An information processing apparatus according to an aspect of the present invention stores a learned model that outputs value information indicating the value of transition to a web page for a web page that can be transitioned in a website. And learning a first acquisition unit that acquires state information indicating a state of a user who refers to a predetermined web page on the website, information indicating the predetermined web page, and state information acquired by the first acquisition unit A second acquisition unit that acquires value information about each of a plurality of transition destination web pages that can be transitioned from a predetermined web page by inputting to a completed model, and a display control unit that controls display of the website Based on the value information acquired by the second acquisition unit, a display system for displaying information for guiding the user to any of the plurality of transition destination web pages on a predetermined web page. It has a part, a.

  ADVANTAGE OF THE INVENTION According to this invention, the technique which can raise a conversion rate can be provided by guide | inducing the user's page transition in a website appropriately.

It is a figure which shows the system configuration example of the web system 1 which concerns on this embodiment. It is a figure which shows the functional block structural example of the server which concerns on this embodiment. It is a figure which shows an example of the process sequence which a learning process part performs. It is a flowchart which shows an example of the process sequence which the server which concerns on this embodiment performs. It is a figure which shows an example of the page transition in the website which the server concerning this embodiment provides. It is a figure which shows an example of the screen displayed on a terminal.

  A preferred embodiment of the present invention will be described with reference to the accompanying drawings. In addition, in each figure, what attached | subjected the same code | symbol has the same or similar structure.

<System configuration>
FIG. 1 is a diagram illustrating a system configuration example of a web system 1 according to the present embodiment. The web system 1 includes a server 10 and a terminal 20. Although one terminal 20 is illustrated in FIG. 1, the web system 1 may include a plurality of terminals 20.

  The server 10 is a server that provides a website. The server 10 may be composed of one or a plurality of physical information processing devices, or may be one or a plurality of virtual information processing devices. Moreover, the server provided by a cloud service may be sufficient.

  The server 10 includes a CPU (Central Processing Unit) 11, a memory, a storage device 12 such as an HDD (Hard Disk Drive) and / or an SSD (Solid State Drive), a communication IF (Interface) 13 for performing wired or wireless communication, and an input operation. An input device 14 for receiving information and an output device 15 for outputting information. The input device 14 is, for example, a keyboard, a touch panel, a mouse, and / or a microphone. The output device 15 is, for example, a display and / or a speaker.

  The terminal 20 is a terminal for referring to a website provided by the server 10, and includes a smartphone, a tablet terminal, a mobile phone, a personal computer (PC), a notebook PC, a personal digital assistant (PDA), a home game device, etc. Any terminal having a communication function capable of communicating with the server 10 can be used.

  As described above, for example, a user who visits a website to conduct a commercial transaction collects information on products and services by referring to various pages in the website, and finally determines whether or not to conduct a commercial transaction. to decide. For example, in the case of an accommodation reservation site, a user who visits a reservation site to make a hotel reservation searches for a candidate hotel based on the location and schedule of the hotel to be reserved, and links to the website. Narrow down the hotels by collecting various information about the candidate hotels while clicking. Eventually, the user makes a hotel reservation if he / she has enough information to make a decision to make a hotel reservation, and if there is not enough information, he / she can make a hotel reservation on the reservation site. I will give up and move to another website. In the present embodiment, reaching a contract is referred to as conversion, and not reaching a contract is referred to as leaving, but the meanings of conversion and leaving may be variously defined according to the purpose of the website.

  Here, what information is required for the user to make a hotel reservation may differ depending on what link the user has followed. For example, a user who arrives at the basic information page A of the hotel A via a link that displays a list of business hotels from the TOP page of the website is assumed. Since this user is assumed to be a user who reserves a hotel for business purposes, it is conceivable that the reservation of the hotel A is decided based on information such as a high discount rate or information such as being close to the station. Therefore, if the user can be guided from the basic information page of the hotel A to a page on which information on price and location is described, the conversion rate (CVR: Conversion Rate) can be increased. It is done.

  Similarly, a user who arrives at the basic information page A of the hotel A via a link that displays a list of hotels near the sightseeing spot from the TOP page of the website is assumed. Since this user is assumed to be a user who reserves a hotel on a trip, it is conceivable to decide on a hotel reservation based on a review that is comfortable or a review that is convenient for sightseeing. Therefore, if the user can be guided from the basic information page of the hotel A to the word-of-mouth page of the hotel A, the conversion rate can be increased.

  Therefore, the server 10 according to the present embodiment uses a learned model generated by performing reinforcement learning using data (such as a link followed by the user) indicating the state of the user who visited the website. The terminal 20 is controlled to display information that guides the user to an effective web page in order to increase the conversion rate. As a result, it is possible to positively provide the user with information necessary for the user to make a business transaction decision, thereby increasing the conversion rate of the website.

<Functional block configuration>
FIG. 2 is a diagram illustrating a functional block configuration example of the server 10 according to the present embodiment. The server 10 includes a storage unit 100, a recording unit 101, a learning processing unit 102, a state information acquisition unit 103 (first acquisition unit), a value information acquisition unit 104 (second acquisition unit), and a Web processing unit 105. And have. The recording unit 101, the learning processing unit 102, the state information acquisition unit 103, the value information acquisition unit 104, and the Web processing unit 105 are realized by the CPU 11 executing a program stored in the storage device 12. be able to. Further, the program can be stored in a storage medium. The storage medium storing the program may be a computer-readable non-transitory storage medium (Non-transitory computer readable medium). The non-transitory storage medium is not particularly limited, but may be a storage medium such as a USB memory or a CD-ROM. The storage unit 100 can be realized using the storage device 12 included in the server 10.

  The storage unit 100 includes a user DB that stores user information of users registered for use on the website, a state information DB that stores information indicating characteristics of users who have accessed the website, and a learning processing unit 102. The learning model (learned model) generated by learning is stored.

  The user DB stores information related to users who are registered for use on the website. For example, the user DB stores user login ID and authentication information, information indicating user attributes (for example, name, age, sex, and / or occupation).

  Each time the website is accessed, the status information DB includes information indicating the status of the user who has accessed the website (hereinafter referred to as “status information”), and whether the user has achieved a predetermined goal on the website. A flag indicating whether or not (hereinafter referred to as “achievement flag”) is stored in association with each other. The status information includes information on the user's flow line in the website (information indicating the transition of all or part of the web page traced from when the user entered the website until conversion or exit). Not limited to. The status information includes, for example, a link source (HTTP referrer) when the user entered the website, information indicating a web page (landing page) that the user first accessed, an attribute of the user who accessed the website, Whether the user is logged in to the website, the month, day, day of the week and / or time when the user accesses the website, the month, day, day of the week and / or time when the user uses the service (for example, staying on weekdays) ) And / or tag information collected on the website may be further included.

  The recording unit 101 has a function of generating state information by using the access log of the website, the user DB, and data acquired from an external analysis server that can communicate with the server 10 and storing the state information in the state information DB. For example, the recording unit 101 may collect status information from an access log or a user DB recorded by a web server function included in the server 10 or inquires of an external analysis server that can communicate with the server 10. Thus, state information may be collected.

  The learning processing unit 102 uses the status information of the user who accesses the website acquired from the status information DB, and information about whether or not the user has achieved a predetermined goal on the website, obtained from the achievement flag. And has a function to perform reinforcement learning. A learned model obtained by performing reinforcement learning is stored in the storage unit 100. Reinforcement learning is a method of learning a learning model so that when an action is given in a certain state, an action that maximizes the reward that can be obtained in the future can be selected as the action to be performed next. . As the reinforcement learning, any learning method can be used, but in the following description, a case where Q learning (Q-learning) is used will be described as an example.

  Q learning is a learning method for learning the value obtained when an action a is given to a certain state s. The obtained value is called a Q value and can be expressed by Q (s, a). The state s in the present embodiment is a state in which a certain user is referring to a certain web page, and is, for example, information on a currently displayed web page and a flow line leading to the web page. Moreover, you may include other state information, such as an attribute of the user who is referring to a web page, in the state s. The action a is an action in which the user moves from the web page to the next web page. The obtained value means that the user finally reaches a conversion (for example, purchasing a product or contracting a service).

  Here, the number of Q (s, a) is equal to the number of combinations of state s and action a. Therefore, if the number that the state s can take and the number that the action a can take become enormous, the number of Q (s, a) also becomes enormous, and it becomes difficult to obtain all Q (s, a) by Q learning. . Therefore, a technique that uses deep learning for expressing Q (s, a) is called DQN (Deep Q Network). By learning a neural network using DQN, a learned model that outputs a Q value by inputting state s and action a can be obtained.

  FIG. 3 shows an example of a processing procedure performed by the learning processing unit 102. First, the learning processing unit 102 sets the state s as the state of the user included in the state information from the state information and the achievement flag stored in the state information DB, and moves the action a from the current web page to the next web page. A combination of learning data generated by setting the reward when the user finally converts to +1 and setting the reward when the user finally leaves the website as −1 is generated (S10). Subsequently, the learning processing unit 102 learns the neural network using the generated learning data (that is, performs learning by DQN), and outputs a Q value when the state s and the action a are input. Is obtained (S11). Returning to FIG.

  The status information acquisition unit 103 has a function of acquiring status information of a user accessing the website. More specifically, the status information acquisition unit 103 acquires status information of a user who is referring to a predetermined web page in the website from a web server function or a user DB included in the server 10.

  The value information acquisition unit 104 inputs information indicating a predetermined web page referred to by the user and the state information acquired by the state information acquisition unit 103 to the learned model, thereby allowing the value information acquisition unit 104 to The Q value (value information) for each of a plurality of transition destination web pages that can be transitioned is acquired. The web page that can be changed from the predetermined web page is not limited to the web page linked from the predetermined web page.

  The web processing unit 105 has a function of displaying various web pages in the website on the terminal 20, a function of accepting a user operation on the web page, and the like. The web processing unit 105 may display a web page on the screen of the terminal 20 and thus may be called a display control unit. In addition, the Web processing unit 105 has a function of passing status information of a user accessing the website at that time to the status information acquisition unit 103. The Web processing unit 105 has a function of outputting an access log of a user accessing the website and passing it to the recording unit 101. Further, the web processing unit 105 has a function of displaying, on a predetermined web page, information that guides the user to any of a plurality of transition destination web pages based on the Q value obtained by the value information obtaining unit 104.

<Processing procedure>
FIG. 4 is a flowchart illustrating an example of a processing procedure performed by the server 10 according to the present embodiment. In the following description, it is assumed that the server 10 provides a website regarding reservation of accommodation facilities. It is assumed that the user accesses the website using the terminal 20 and refers to various web pages on which information on accommodation facilities is posted in order to make a reservation for a desired accommodation facility.

  First, the status information acquisition unit 103 inquires of the Web processing unit 105 to acquire information indicating a web page that is being referred to by the user (for example, a URL or an ID that identifies the web page). Further, the status information acquisition unit 103 acquires the status information of the user who is referring to the web page by analyzing, for example, an access log output by the web processing unit 105 (S20). Moreover, when the user who is referring to the website is logged in to the website, the status information acquisition unit 103 acquires the user attribute as part of the user status information by accessing the user DB.

  Subsequently, the value information acquisition unit 104 inputs information indicating the web page referred to by the user and the state information of the user into the learned model, so that each web page of the next link destination (transition destination) Q value is acquired (S21). Subsequently, the web processing unit 105 displays information recommending a transition to a linked web page having the largest Q value on the web page referred to by the user (S22). When a plurality of users are browsing the website, the processing procedure from step S20 to step S22 is executed for each user every time the user transitions to one of the web pages.

  FIG. 5 is a diagram illustrating an example of page transition in a website provided by the server 10 according to the present embodiment. For example, it is possible to transition from the page 1-1 to the page 1-100 from the top page, and further transition from the page 1-99 to any of the pages 2-50 to 2-53. it can. Similarly, the page 2-50 can transition to any of page 3-20 to page 3-23. When a contract is finally reached, the page transitions to a CV page (for example, a page that presents a reservation number or the like to the user).

  The learned model is learned so as to output a Q value for each next transition destination web page for all pages shown in FIG. For example, assume that the user refers to the web page in the order of the top page, pages 1-99, and pages 2-50. In this case, the web page referred to by the user is 2-55, and the page transition from the top page to the web page 2-55 (top page, page 1-99, page 2-50 has been made in this order). When the user state information including information indicating that is input to the learned model, the learned model outputs a Q value for each of page 3-20 to page 3-23. Here, it is assumed that the Q value for page 3-23 is the largest among the output Q values.

  FIG. 6 is a diagram illustrating an example of a screen displayed on the terminal 20. FIG. 6A, FIG. 6B, and FIG. 6C show an example of the top page, an example of the page 1-99, and an example of the page 2-50, respectively. In FIG. 6C, a menu button M10 for displaying a menu related to the inn A is provided. When the menu button M10 is pressed (touched), the link to the next transition destination web page is “Price / A link to a page displaying "Accommodation plan", a page displaying "Photo / Movie", a page displaying "Map / Access method", and a page displaying "Review" is displayed. A page displaying “Price / Accommodation Plan”, a page displaying “Photo / Video”, a page displaying “Map / Access Method”, and a page displaying “Review” are shown on page 3-20 in FIG. , Page 3-21, page 3-22, and page 3-23.

  As described above, the Q value of page 3-23 is the largest among the Q values output from the learned model. This indicates that when the user refers to a page displaying “reviews” next to the page shown in FIG. 6C, there is a high possibility that the user will eventually contract an accommodation facility.

  In this case, as illustrated in FIG. 6C, the web processing unit 105 displays, on the web page, information that guides the user to the transition destination web page having the largest Q value among the plurality of transition destination web pages. The information for guiding the user may be information for prompting the user to perform a predetermined action, and may be a pop-up displayed on the web page as shown in FIG. The pop-up may include link information on the “review” page. It is possible to increase the probability that the user refers to the “review” page. The information for guiding the user is not limited to the pop-up. For example, the web processing unit 105 may not display a menu other than “review” on the menu M10. Further, for example, the Web processing unit 105 may transmit a push notification including a message that prompts the terminal 20 to refer to the review page.

<Modification>
When there is a user who has converted, the learning processing unit 102 may further learn the learned model using the state information newly added to the state information DB regarding the user. In the learning, a user who has not been converted may be included in the learning target. It is possible to improve the accuracy of the Q value automatically output from the learned model.

  In the embodiment described above, the Web processing unit 105 displays information for guiding the user to the transition destination web page having the largest Q value on the web page. However, the present invention is not limited to this. For example, the web processing unit 105 uses, as a web page, information that guides the user to at least one of the transition destination web pages having a Q value equal to or higher than a predetermined rank among the plurality of transition destination web pages. You may make it display.

  For example, it is assumed that the Q value is large in the order of a page displaying “Price / Accommodation Plan”, a page displaying “Photo / Video”, a page displaying “Map / Access Method”, and a page displaying “Review”. . Further, it is assumed that the predetermined rank or higher is second or higher. In this case, the Web processing unit 105 displays either one or both of the pop-up described as “view map / access method” and the pop-up described as “view review” on the screen of FIG. You may make it display on.

  As a result, in the state information newly accumulated in the state information DB, the transition of the web page followed by the user is biased to a specific pattern, and the accuracy of the learned model is not improved when the learned model is further learned. It is possible to suppress this possibility.

<Summary>
According to the embodiment described above, it is possible to appropriately guide the user's page transition on the website, and to increase the conversion rate. In addition, by appropriately guiding the page transition to the user, it is possible to suppress reference to a page that is difficult to improve the conversion rate, and to reduce the amount of data transmitted and received between the server 10 and the terminal 20. It becomes possible.

  The embodiments described above are for facilitating the understanding of the present invention, and are not intended to limit the present invention. The flowcharts, sequences, and elements included in the embodiments, and the arrangement, materials, conditions, shapes, sizes, and the like included in the embodiments are not limited to those illustrated, but can be changed as appropriate. In addition, the structures shown in different embodiments can be partially replaced or combined.

  DESCRIPTION OF SYMBOLS 10 ... Server, 11 ... CPU, 12 ... Storage device, 13 ... Communication IF, 14 ... Input device, 15 ... Output device, 20 ... Terminal, 100 ... Storage unit, 101 ... Recording unit, 102 ... Learning processing unit, 103 ... Status information acquisition unit, 104 ... value information acquisition unit, 105 ... Web processing unit

Claims (8)

A storage unit that stores a learned model that outputs value information indicating the value of transition to the web page for a web page that can be transitioned in the website;
A first acquisition unit that acquires state information indicating state of a user who refers to a predetermined web page on the website and includes information on transition of the web page followed by the user on the website ;
Each of a plurality of transition destination web pages that can transition from the predetermined web page by inputting information indicating the predetermined web page and the state information acquired by the first acquisition unit to the learned model. A second acquisition unit for acquiring value information about
A display control unit that controls display of the website, wherein information that guides the user to any of the plurality of transition destination web pages based on the value information acquired by the second acquisition unit A display control unit to be displayed on the web page;
An information processing apparatus. The display control unit sends information that guides the user to the transition destination web page having the largest value indicated by the value information acquired by the second acquisition unit among the plurality of transition destination web pages. Display,
The information processing apparatus according to claim 1. The display control unit may select at least one of the plurality of transition destination web pages as a transition destination web page whose value indicated by the value information acquired by the second acquisition unit is greater than or equal to a predetermined rank. Displaying information for guiding the user on the predetermined web page;
The information processing apparatus according to claim 1. The information for guiding the user is a popup displayed on the predetermined web page.
The information processing apparatus according to any one of claims 1 to 3 . The information for guiding the user includes link information to a transition destination web page for guiding the user.
The information processing apparatus according to any one of claims 1 to 4 . The learned model is a learned model generated by reinforcement learning based on status information of a user who has accessed the website and information on whether or not the user has achieved a predetermined goal on the website. ,
The value information is a Q value.
The information processing apparatus according to any one of claims 1 to 5 . On the computer,
A storage unit that stores a learned model that outputs value information indicating the value of transition to the web page for a web page that can be transitioned in the website;
A first acquisition unit that acquires state information indicating state of a user who refers to a predetermined web page on the website and includes information on transition of the web page followed by the user on the website ;
Each of a plurality of transition destination web pages that can transition from the predetermined web page by inputting information indicating the predetermined web page and the state information acquired by the first acquisition unit to the learned model. A second acquisition unit for acquiring value information about
A display control unit that controls display of the website, wherein information that guides the user to any of the plurality of transition destination web pages based on the value information acquired by the second acquisition unit A display control unit to be displayed on the web page;
A program for running A display control method performed by the information processing apparatus,
Storing, in a storage unit, a learned model that outputs value information indicating the value of transitioning to a web page for a web page that can be transitioned in a website;
Obtaining state information including state information indicating a state of a user who refers to a predetermined web page on the website and information on a transition of the web page followed by the user on the website ;
By inputting information indicating the predetermined web page and the acquired state information into the learned model, value information about each of a plurality of destination web pages that can be transitioned from the predetermined web page is acquired. Steps,
A display control unit that controls display of the website, the step of causing the predetermined web page to display information for guiding the user to any of the plurality of transition destination web pages based on the acquired value information When,
A display control method including: