JP6535564B2 - INFORMATION PROCESSING APPARATUS, INFORMATION PROCESSING METHOD, AND PROGRAM - Google Patents

Description

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

There is a technique of analyzing a user's behavior tendency using data indicating user attribute information such as user's (for example, website browsing person) behavior history and preference information in response to an analysis request from an analyst. The action history is, for example, a browsing history of a website, or a purchase history of a product via the website. In addition, there is a technique of presenting an analysis result indicating an action tendency of a user group having specific user attribute information to an analyst based on an analysis result of the action tendency. Then, based on the obtained analysis result, the analyst performs various operations that are the purpose of analysis, for example, selection of a distribution target user to which an advertisement is to be distributed.
For example, the product purchase data processing system described in Patent Document 1 analyzes the purchase tendency according to the customer attribute from the purchase history of the customer having the specific attribute.

JP 2001-216369 A

  By the way, decision tree analysis may be used as an analysis method for analyzing a user's action tendency. The decision tree analysis calculates a decision tree represented in a data format of a tree structure from data indicating user attribute information and an action history of a user to be analyzed. A decision tree has a tree structure of nodes representing classifications and rules (branches) representing a collection of features leading to the classifications. Decision tree analysis is an analysis method that can classify users by calculating a decision tree and visualize and present analysis result data indicating behavior tendencies for each classified user. The analysis result of such decision tree analysis may be used for decision support for the service provision contents and provision method for the user.

The analyst selects a node that is determined to be characteristic in the calculated decision tree, and performs a task (for example, advertisement delivery) targeted by the analysis for the user group belonging to the classification indicated by the selected node. I do. Here, it is preferable that the selection of nodes be made to have a large effect on the task of analysis. The effects include, for example, acquisition of new users, improvement of revisit rates of existing users, or expansion of sales.
However, when using data indicating the user attribute information of the user to be analyzed and the action history, etc. for analysis, it is grasped if the data within which period is used for analysis will increase the effect on the task of analysis purpose. It is not easy. For example, whether analysis using data for the last one month is effective or analysis using data for the last one year is effective is the target of the operation or analysis that is the objective of the analysis. It depends on the type of industry.

  The present invention has been made in view of the above points, and is an information processing apparatus capable of presenting an analysis result in which a feature is more strongly expressed among a plurality of analysis results having different analysis target periods, an information processing method, And provide programs.

(1) The present invention has been made to solve the above-described problems, and one aspect of the present invention is to obtain data indicating user attributes of each of a plurality of users, and to obtain data during a specific period of the data. A decision tree calculating unit that calculates a plurality of decision trees that respectively classify the user into any one of a plurality of nodes based on the user attribute of the corresponding data, and the user classified into the nodes A plurality of decision trees calculated by the decision tree calculation unit on the basis of a representative index calculation unit that calculates values of representative indexes that are indices based on numbers, and the values of the representative indexes calculated by the representative index calculation unit A verification evaluation unit that selects any decision tree from among the data, data indicating the decision tree selected by the verification evaluation unit, and data indicating the value of the representative index calculated by the representative index calculation unit An output unit for outputting the analysis result information including the decision tree is calculated based on the data, an information processing apparatus comprising: a.

(2) Further, as one aspect of the present invention, the verification evaluation unit distributes the value of the representative index for each of the nodes calculated by the representative index calculation unit with respect to the decision tree for each of the plurality of different periods. The information processing apparatus according to (1), wherein a decision tree having features represented by values of the representative indexes is selected by calculating each and selecting the decision tree with the largest variance. .

(3) Further, according to an aspect of the present invention, the representative index is a ratio of the number of users who acted as an analysis target among the number of users classified into the nodes ( It is an information processor given in 1) or (2).

(4) Further, according to an aspect of the present invention, the representative index is a user classified into the node in a ratio of the number of users who perform an analysis target among the number of users classified into the node. The information processing apparatus according to any one of (1) to (3), which is a value obtained by multiplying the average sales per person.

(5) Further, according to an aspect of the present invention, in the information processing method using a computer, the decision tree calculation unit included in the computer acquires data indicating user attributes of each of a plurality of users, and the data A decision tree calculating step of computing a plurality of decision trees each of which classifies the user into any one of a plurality of nodes based on the user attribute of data corresponding to a specific period among the plurality of nodes; A representative index calculation step of calculating a representative index value calculated based on the number of users classified into the nodes, and a verification evaluation unit included in the computer; From among the plurality of decision trees calculated by the decision tree calculating step based on the value of the representative index calculated in A verification and evaluation step of selecting one of the decision trees, and an output unit of the computer indicates data indicating the decision tree selected in the verification and evaluation step and the value of the representative index calculated in the representative index calculation step An output step of outputting analysis result information indicating a decision tree calculated on the basis of data.

(6) Further, according to an aspect of the present invention, the computer acquires data indicating user attributes of a plurality of users, and the user of the plurality of pieces of data is determined based on the user attributes of data corresponding to a specific period. Decision tree calculating step of calculating a plurality of decision trees each of which is classified into any one of a plurality of nodes by changing the period, and a value of a representative index calculated based on the number of the users classified into the nodes Based on the representative index calculating step to be calculated and the value of the representative index calculated in the representative index calculating step, any decision tree is selected from among the plurality of decision trees calculated in the decision tree calculating step Verification evaluation step, data indicating the decision tree selected in the verification evaluation step, and the representative finger calculated in the representative index calculation step An output step of outputting the analysis result information indicating a decision tree which is calculated based on the values in the data indicating a program for execution.

  According to the present invention, it is possible to present an analysis result with more distinctive features among a plurality of analysis results with different analysis target periods.

BRIEF DESCRIPTION OF THE DRAWINGS It is the schematic which shows the outline | summary of the information processing system which concerns on this embodiment. It is a block diagram showing functional composition of an information processor concerning this embodiment. It is a figure which shows an example of the node selection performance table stored in the undercut value storage part of the information processing apparatus which concerns on this embodiment. It is a figure which shows an example of the image output by the output part of the information processing apparatus which concerns on this embodiment. It is a flowchart which shows operation | movement of the information processing apparatus which concerns on this embodiment.

(Embodiment)
Hereinafter, an outline of the information processing system according to the present embodiment will be described with reference to the drawings.
FIG. 1 is a schematic view showing an outline of an information processing system according to the present embodiment.
The information processing system according to the present embodiment is a Web advertisement distribution support system for supporting the decision making of a person in charge of distribution of a Web advertisement (hereinafter referred to as Web advertisement). The information processing system supports the decision making of a person in charge of web advertisement distribution (hereinafter, referred to as an analyst) in selecting a web advertisement distribution target user.

As illustrated, the information processing system includes an information processing apparatus 1, a web server 2, and a web advertisement distributor terminal 3.
The information processing device 1 is a Web advertisement distribution support device that supports the analyst's decision making in selecting a Web advertisement distribution target user. The information processing apparatus 1 is installed, for example, on the premises of a company to which an analyst belongs. The information processing apparatus 1 is configured to include, for example, a personal computer or a general-purpose computer.

The web server 2 is a server device that publishes a web site via a communication network (for example, the Internet). The website is, for example, an EC (Electronic Commerce) site. Although only one web server 2 is shown in FIG. 1 to simplify the description, the number of web servers 2 may be plural.
In addition to a server device that publishes a website operated by a company to which the analyst belongs (hereinafter also referred to as an in-house website), a plurality of web servers 2 are operated by a company other than the company to which the analyst belongs. A server apparatus that publishes a website (hereinafter also referred to as a third-party website) may be included.

  The web advertisement distributor terminal 3 is a terminal device that distributes the web advertisement to the user according to the execution instruction of the web advertisement distribution based on the data input from the information processing device 1. The web advertisement distributor terminal 3 is a terminal device installed on the premises of the web advertisement distributor and in communication connection with the information processing apparatus 1 through a communication network. The web advertisement distributor is, for example, a DSP (Demand-Side Platform) that operates a demand-side platform (a platform for supporting advertising effectiveness maximization on the advertiser side in online advertising). The web advertisement distributor terminal 3 includes, for example, a personal computer or a general-purpose computer.

  The information processing apparatus 1 acquires, from the web server 2 that publishes the company website, log data indicating the browsing history of the company website for each viewer (hereinafter, also referred to as a user) of the company website. Further, the information processing apparatus 1 acquires, from the web server 2 that publishes the web site of the other company, log data indicating the browsing history of the web site of the other company for each user.

In response to an analysis request based on data input by the analyst, the information processing device 1 analyzes the user's behavior tendency. The content of the analysis request from the analyst is, for example, requesting analysis of the behavior tendency of the user who has purchased a specific product (for example, the product A).
The information processing apparatus 1 uses the user attribute information of the user based on the log data indicating the browsing history of the company website for each user acquired from the Web server 2 and the log data indicating the browsing history of the other company website for each user. Analyze The user attribute information includes preference information and outcome information.

  The preference information is data indicating things that the user is interested in, for example, data indicating a genre (for example, sports, movies, etc.) included in the content of the web site that the user browses. Further, the result information is data indicating the user's behavior on the website, for example, data indicating whether a product purchase has been made or whether a material request has been made.

The information processing apparatus 1 outputs analysis result data indicating a user group in which the user is classified for each user attribute based on the user attribute information. For example, the information processing apparatus 1 classifies the web page belonging to the “science” genre into a user group in which the number of times the web page belonging to the “science” category is 20 or more and a user group in which the web page is less than 20 times. Then, the information processing apparatus 1 analyzes, for example, the ratio of users who have purchased the product A for each user group, and outputs analysis result data.
The information processing apparatus 1 according to the present embodiment classifies users using a decision tree analysis method, and the details will be described later.

The information processing apparatus 1 receives an input of data indicating a user group selected by the analyst from among a plurality of user groups included in the analysis result data. The information processing device 1 transmits data indicating the user group to the Web advertisement distributor terminal 3.
The web advertisement distributor terminal 3 distributes the web advertisement to the terminal of the user included in the user group based on the data input from the information processing device 1.
In the present embodiment, the web advertisement distributor terminal 3 directly delivers the web advertisement to the user's terminal, but the present invention is not limited to this. For example, the Web advertisement distributor terminal 3 may be configured to distribute an advertisement to a user via an electronic mail. That is, the web advertisement distributor terminal 3 describes an e-mail in which an advertisement is described or a URL of a web page on which a web advertisement is posted (Uniform Resource Locator; uniform resource locator) to the user's e-mail address. It may be configured to send an e-mail or the like.

(Configuration of information processing apparatus)
Hereinafter, the functional configuration of the information processing apparatus 1 according to the embodiment of the present invention will be described in detail with reference to the drawings.
FIG. 2 is a block diagram showing a functional configuration of the information processing apparatus 1 according to the present embodiment.
The information processing apparatus 1 includes a control unit 10, an input unit 11, a history storage unit 12, a site preference definition unit 13, a simulation verification evaluation unit 14, a decision tree calculation unit 15, and a representative index calculation unit 16. The threshold value storage unit 17, the output unit 18, the advertisement distribution unit 19, and the display method storage unit 20 are included.

The control unit 10 controls various processes of the information processing apparatus 1. The control unit 10 includes, for example, a central processing unit (CPU).
The input unit 11 receives an analysis request input by the analyst. The input unit 11 outputs data indicating the input analysis request to the control unit 10. The input unit 11 includes, for example, a mouse, a keyboard, or a touch panel.

  The history storage unit 12 stores log data indicating a company website browsing history for each user. The history storage unit 12 acquires, from the web server 2, log data indicating a company website browsing history. The company website referred to here is a website published by the web server 2 via the communication network. The web server 2 is, for example, a web server of an EC site operated by a company to which an analyst belongs. Also, in the company website browsing history, the name of the web page accessed in the company website, the date and time of access, and the result (for example, whether the product has been purchased or whether a document request has been made) Data indicating the

  Further, the history storage unit 12 stores log data indicating the other company website browsing history for each user. The history storage unit 12 acquires, from the Web server 2, log data indicating the other company's website browsing history, for example, by referring to referrer information stored in the Web server 2. The other company's Web site mentioned here is a Web site operated by another company (that is, for example, a company other than the company to which the analyst belongs). The other-company Web site browsing history includes the name of the Web page accessed in the other-company Web site based on the cookie (Cookie) information of the accessed user, the access date, and the like.

  The history storage unit 12 may be a storage medium such as a hard disk drive (HDD; Hard Disk Drive), a flash memory, an EEPROM (Electrically Erasable Programmable Read Only Memory), a ROM (Read Only Memory), a RAM (Read Only Memory), or the like. Random Access read / write Memory (Read / Write Memory), or any combination thereof.

The site preference definition unit 13 associates and stores in advance attribute information of web pages in another company's web site and information for identifying a web page in another company's web site.
The attribute information of the web page in another company's web site is, for example, data indicating the genre (for example, sports, movie) of the content included in the web site. The attribute information of the web page in the other company website is utilized as preference information of the user in decision tree analysis.
Further, the information for identifying the web page in the other company's web site is, for example, the name or the URL of the web page in the other company's web site.

  The site preference definition unit 13 is configured to include a storage medium, for example, a hard disk drive, a flash memory, an Eyprom, a ROM, a RAM, or any combination thereof.

  The history storage unit 12 associates the identification information of the web page accessed by the user in the other company website with the identification information of the web page in the other company website stored in the site preference definition unit 13. As a result, the history storage unit 12 can acquire attribute information of the web page accessed by the user in the web site of another company. That is, the history storage unit 12 can acquire, for example, data indicating the genre to which the content of the web page in the other company website accessed by each user belongs.

When data indicating an analysis request from the analyst is input from the input unit 11, the control unit 10 outputs the data to the simulation verification evaluation unit 14.
When data indicating an analysis request from the analyst is input from the control unit 10, the simulation verification and evaluation unit 14 receives user attribute information for each user, that is, result information and preference information (for example, another company's website viewed by the user) The log data indicating the genre of the web page, etc.) is acquired from the history storage unit 12.

  The simulation verification evaluation unit 14 extracts log data recorded within a specific period from the log data acquired from the history storage unit 12. The simulation verification evaluation unit 14 outputs the extracted log data to the decision tree calculation unit 15 via the control unit 10, and causes the decision tree calculation unit 15 to calculate a decision tree based on the log data. The simulation verification and evaluation unit 14 acquires data indicating the decision tree calculated by the decision tree calculation unit 15 via the control unit 10.

The simulation verification and evaluation unit 14 performs the extraction of the log data recorded in the specific period as described above in the same manner for a plurality of periods while changing the period.
The plurality of periods to be extracted from the log data are, for example, a plurality of periods having different lengths, such as “last week”, “last month”, and “last year”. Alternatively, the plurality of periods to be extracted from the log data may be, for example, “period from 10 days to yesterday,” “period from 20 days to 11 days,” and “period from 30 days to 21 days”. As such, there may be a plurality of periods where the beginning and end of the period are different. Alternatively, the plurality of periods targeted for extraction of the log data may be a combination of a plurality of periods having different lengths of periods and a plurality of periods having different beginnings and ends of the periods.

  The simulation verification and evaluation unit 14 outputs the extracted log data of a plurality of periods to the decision tree calculation unit 15 through the control unit 10, and the decision tree calculation unit 15 determines a decision tree based on the log data of a plurality of periods. Let each calculate. The simulation verification and evaluation unit 14 acquires data indicating the decision tree of each of the plurality of periods calculated by the decision tree calculation unit 15 through the control unit 10.

The decision tree calculation unit 15 performs decision tree analysis based on the result information and preference information for each user input from the simulation verification and evaluation unit 14 via the control unit 10, and calculates a decision tree. The calculated decision tree includes, for each node, values of “number”, “ratio”, and “occupancy (occupancy ratio)”. “Number” represents the number of users corresponding to the condition corresponding to each node. "Rate" represents the ratio of each segment in the condition corresponding to each node. "Saiyou" represents the volume of a segment corresponding to the condition corresponding to each node.
The simulation verification and evaluation unit 14 outputs, to the representative index calculation unit 16 via the control unit 10, data indicating the decision trees for each of the plurality of periods acquired from the decision tree calculation unit 15 via the control unit 10.

  The representative index calculation unit 16 calculates a representative index for each node of a decision tree for each of a plurality of periods based on data input from the simulation verification and evaluation unit 14 via the control unit 10. The representative index indicates the size of the expectation when setting the purpose. In the present embodiment, the representative index is assumed to be the value of “ratio” at each node, but is not limited to this. The representative index is preferably an index determined by an appropriate calculation method according to the purpose of analysis.

For example, if the purpose of analysis is to improve the product purchase rate, then the value itself of the “ratio” at each node is used as the representative index, for example. That is, the representative index in this case is the expected value of the product purchase rate of the user included in each node.
Also, for example, if the purpose of analysis is to expand expected sales, the representative index is, for example, a value obtained by multiplying the “ratio” at each node by the average sales per user of users included in the node. Is used. That is, the representative index in this case is the expected value of average sales per person of the user included in each node.
In addition, according to the purpose of analysis, the representative user uses the product repeat purchase rate, the website visit time, the number of web page visits, or the interval from the previous product purchase (visit) as a representative index for users included in each node It may be done.

The representative index calculation unit 16 performs simulation verification and evaluation unit via the control unit 10 with data indicating the value of the representative index of each node of the decision tree for each of a plurality of periods, and data indicating the decision tree for each of the plurality of periods. Output to 14.
The simulation verification and evaluation unit 14 selects a decision tree in which the feature of the value of the representative index appears most clearly, based on the value of the representative index of each node of the decision tree for each of a plurality of periods. For example, the simulation verification and evaluation unit 14 calculates the variance of the value of the representative index of each node of the decision tree for each decision tree of each of a plurality of periods. Then, the simulation verification and evaluation unit 14 selects a decision tree with the largest variance based on the calculated result.
The simulation verification and evaluation unit 14 outputs data indicating the selected decision tree to the representative index calculation unit 16 via the control unit 10.

  The threshold value storage unit 17 stores, for example, the value of the “number” of nodes selected by a skilled analyst in past analysis (hereinafter referred to as a threshold value). The cut-off value storage unit 17 is configured to include a storage medium, such as a hard disk drive, a flash memory, an easy prom, a ROM, a RAM, or any combination thereof.

FIG. 3 is a diagram showing an example of the node selection result table stored in the footing value storage unit 17 of the information processing apparatus 1 according to the present embodiment.
As shown, the node selection result table is a two-dimensional tabular data consisting of two columns of "analyst's skill level" and "number of users included in the selected node". is there. Each row of the node selection result table corresponds to the result of node selection by the analyst in the past.

For example, data “high” and “54” are stored in the top (uppermost) data row in the node selection result table shown in FIG. This indicates that the skill level of the analyst who made the node selection indicated by this data row is “high” (an analyst with a high skill level). It also indicates that the number of users included in the node selected in the node selection indicated by this data row is 54.
Also, for example, data “standard” and “42” are stored in the second data row in the node selection result table shown in FIG. This indicates that the skill level of the analyst who made the node selection indicated by this line is "standard" (the analyst of the standard skill level). It also indicates that the number of users included in the node selected in the node selection indicated by this data row is 42.

  The representative index calculation unit 16 is a value below the value of “number” of nodes selected by the skilled analyst for the value of “number” in each node of the decision tree selected by the simulation verification evaluation unit 14 above. Then, the value of the representative index of the node is set to 0. That is, if the number of users at each node is less than the threshold value, the value of the representative index is zero.

  Note that up to which skill level analysts are "skilled analysts" (for example, only "skill level" analysts are "skilled analysts" or "skill level" is "high") Whether the analyst and the analyst having the skill level “standard” are combined to be “skilled analyst”, etc. is set in advance.

  Essentially, the higher the response rate from the user can be expected, the more efficiently the advertisement can be distributed, as the advertisement is distributed to the user included in the node having the larger value of “ratio” which is the representative index. However, even if the value of the “ratio” of the node selected by the analyst is large, the number of users (volumes) who respond to the distributed advertisement if the number of users included in the node is too small. Not too many. It is not efficient to deliver advertisements to such a group of users with too few users.

However, at least how many users are effective in delivering the advertisement depends, for example, on the content of the advertisement (for example, the product to be advertised). Therefore, the decision to determine if the number of users delivering the ad is likely to be effective ad delivery can be made based on the ad delivery experience by a skilled analyst, etc. preferable.
For the above reason, the representative index calculation unit 16 of the information processing apparatus 1 according to the present embodiment determines nodes with a small number of users according to the “threshold value” determined based on the judgment in the past analysis of the skilled analyst. Cut short. The information processing apparatus 1 makes the node an analyst by making the output contents of the decision tree (the display state of the node, for example, the density of the color in the frame of the node) different according to the cutoff. Make it hard to select. An output example of the decision tree will be described later.

The control unit 10 is data indicating the decision tree calculated by the decision tree calculation unit 15, the value of the representative index of each node calculated by the representative index calculation unit 16, and the display method stored in the display method storage unit 20 described later. Based on the information, data (analysis result information) indicating an image visualizing the decision tree is generated, and the output unit 18 outputs the image.
The output unit 18 is configured to include a display, for example, a liquid crystal display, or an organic EL (Electro Luminescence) display.

(Example output of decision tree)
Below, an example of the image of the decision tree which the output part 18 outputs is demonstrated.
FIG. 4 is a view showing an example of an image output by the output unit 18 of the information processing apparatus 1 according to the present embodiment.
As illustrated, the decision tree dt1 in this example includes thirteen nodes (node nd0, node nd1, node nd2,..., Node nd12). The 12 nodes (node nd1, node nd2, node nd3,..., Node nd12) except for the node nd0 of the top layer are the nodes above the first layer and the rules (rule ed1, rule ed2, rule ed3,. · · · Are connected by the rule ed12). Thus, the decision tree dt1 has a tree structure.

In the lower part of the node nd0 which is a node of the top hierarchy, "whole" is displayed. The words displayed at the bottom of each node represent classification conditions for classifying data to be analyzed. That is, the node nd0 is a node that represents the result of counting for all users to be analyzed.
In the upper part of the frame of the node nd0, “number: 99000: 1000” is displayed. This classifies the analysis target user according to the presence or absence of the action target user (in the present example, the presence or absence of the purchase history of the product A), and represents the number of users corresponding to the presence or absence of each action. In this example, the numbers on the left represent the number of users without a purchase history, and the numbers on the right represent the number of users with a purchase history. That is, the upper number in the frame of the node nd0 indicates that 99000 have no purchase history of the product A and 1000 have a purchase history of the product A. As can be seen from these figures, the total number of analysis target users in this example is a total value of 99,000 and 1,000, and is 100,000.

  In the lower part of the frame of the node nd0, “ratio: 99.0: 1.0” is displayed. This classifies the analysis target users according to the presence or absence of the action target user (in the present example, the presence or absence of the purchase history of the product A), and represents the ratio of the number of users corresponding to the presence or absence of each action. In this example, the number on the left represents the proportion of users without a purchase history, and the number on the right represents the proportion of users with a purchase history. That is, the middle number in the frame of node nd0 indicates that 99.0% of users included in node nd0 (that is, all users to be analyzed) have no purchase history of product A, and 1.0% The user indicates that there is a purchase history of the product A.

  The node nd0 is connected to a node nd1 displayed in the lower left direction of the node nd0 by a rule eg1. The node nd0 is connected to a node nd2 displayed in the lower right direction of the node nd0 by a rule eg2. The nodes nd1 and nd2 are nodes corresponding to a hierarchy one stage lower than the node nd0.

Below the node nd1, "Science <20" is displayed. As described above, the words displayed at the bottom of each node represent classification conditions for classifying data to be analyzed. That is, the node nd1 is a node representing the result of counting for users corresponding to “science <20” among all users to be analyzed. In the present example, the classification condition represented by “science <20” is assumed to be the classification condition based on the number of times of reference to the web page related to science. That is, the node nd1 is a node representing the result of counting for users whose number of times of reference to the Web page related to the science is less than 20 among the users to be analyzed.
Similarly, “Science> = 20” is displayed below the node nd2. That is, the node nd1 is a node representing the result of aggregation for users whose number of times of reference to the Web page related to science is 20 or more among all users to be analyzed.

  In this manner, the decision tree dt1 indicates two groups of users included in a certain node according to a specific classification condition (for example, whether or not the number of times the Web page related to science is referenced is 20 or more). It is a decision tree that classifies and displays the nodes corresponding to each classified user group as nodes of a hierarchy one level lower.

  In the upper part of the frame of the node nd1, “number: 98992: 998” is displayed. This classifies the users included in the node nd1 (that is, users among the analysis target users who have referred to the Web page related to science less than 20 times) by the presence or absence of the purchase history of the product A, and The number of users corresponding to the presence or absence of the action of That is, the upper number in the frame of the node nd1 indicates that among the users included in the node nd1, 98992 have no purchase history of the product A and 998 have a purchase history of the product A. As can be seen from these figures, the number of users included in the node nd1 (ie, among the analysis target users who have referred to a Web page related to science less than 20 times) is 98992 It is the total value of 998 people and is 99999 people.

  In the middle part of the frame of the node nd1, “ratio: 99.0: 1.0” is displayed. This classifies the users included in the node nd1 according to the presence or absence of the purchase history of the product A, and represents the ratio of the number of users corresponding to the presence or absence of each action. That is, the middle number in the frame of the node nd1 is 99 of the users included in the node nd1 (that is, users among the analysis target users who have referred to a Web page related to science less than 20 times) 9% indicates that there is no purchase history of product A, and 1.0% indicates that there is a purchase history of product A.

  In the lower part of the frame of the node nd1, it is displayed as “Shingo: 100.0: 99.8”. The number “100.0”, which is the number on the left, becomes the analysis target in the user included in the node nd1 in the entire user who does not have the activity of the user to be analyzed (in this example, the purchase of the product A). It represents the occupancy rate of the user without the user's action. That is, “100.0”, which is the number on the left side, indicates that the purchase history of the product A by the user having the number of times of referring to the Web page related to science is less than 20 It represents the occupancy rate of non-users.

  As described above, the total number of users who do not have a purchase history of product A is 99000, and the user who has not browsed a Web page related to science less than 20 times does not have a purchase history of product A Since the number of is 98992 people, the occupancy rate is “99.919 ...%”, but the decision tree dt1 in this example displays up to one decimal place. As a result, “100.0%”, which is a value obtained by rounding “99.9919 ···%”, is displayed from the lower part of the frame of the node nd0.

  Similarly, “99.8”, which is the lower right number in the frame of the node nd1, indicates the behavior of the user to be analyzed (purchase of the product A in this example) in the entire analysis target user. In the total number of users, the occupancy rate of the user having the user's behavior to be analyzed among the users included in the node nd1 is represented. That is, “99.8”, which is the number on the right, indicates that the purchase of product A by a user who has less than 20 times of reference to a science related web page in the total number of users who have a purchase history of product A It represents the occupancy rate of the number of users who have a history.

  Next, “number: 8: 2” is displayed in the upper stage within the frame of the node nd2. That is, the upper number in the frame of the node nd2 indicates that, among the users included in the node nd2, eight have no purchase history of the product A and two have a purchase history of the product A. As can be understood from these figures, the number of users included in the node nd2 is a total value of eight and two, and is ten.

  In the middle part of the frame of the node nd2, “ratio: 80.0: 2.0” is displayed. That is, the middle number in the frame of the node nd2 is 80 of the users included in the node nd2 (that is, users among the analysis target users who have referred to a Web page related to science less than 20 times) .0% indicates that there is no purchase history of product A, and 20.0% indicates that there is a purchase history of product A.

  In the lower part of the frame of the node nd2, it is displayed as "Shingo: 0.0: 0.2". That is, the number “0.0”, which is the number on the left, corresponds to the number of users who have referred to the science related Web page in the total number of users who do not have a purchase history of the product A at 20 or more times. It represents the occupancy rate of the number of users who have no purchase history.

  As described above, the total number of purchased users of product A is 99000, and the number of users who purchased product A is 8 with 20 or more times referring to the Web page related to science Although the occupancy rate is “0.0080...%”, The rounding-off value “0.0%” is displayed.

  Similarly, the lower rightmost digit “0.2” in the frame of the node nd 2 indicates that the number of times the science related web page has been referred to in the total number of users who have a purchase history of the product A is 20. It represents the occupancy rate of the number of users who have a purchase history of the product A for the user who is more than the number of times.

  Next, “number: 83682: 758” is displayed in the upper part of the frame of the node nd3, which is a node in a hierarchy one step lower than the node nd1. That is, the upper number in the frame of the node nd3 indicates the user included in the node nd3 (that is, among the analysis target users, the number of times the Web page related to science has been referenced less than 20 times and related to health Among the users who have referred to the Web page to be accessed less than once (that is, users who do not refer to it), there are 83,682 people who do not have a purchase history of product A and 758 have a purchase history of product A Show. As can be seen from these figures, the number of users included in the node nd3 is the sum of 8,3682 and 7,758, and is 84,440.

  In the middle part of the frame of the node nd3, “ratio: 99.1: 0.9” is displayed. That is, the middle number in the frame of the node nd3 indicates that among the analysis target users, the number of times of referring to the Web page related to science is less than 20 and the number of times of reference to the Web page related to health is 1 Among the users who are less than the number of times (that is, not referred to), the 99.1% user indicates that there is no purchase history of product A, and the 0.9% user has a purchase history of product A.

  In the lower part of the frame of the node nd3, it is displayed as "Shingo: 84.5: 75.8". That is, the number on the left side of “84.5” refers to the number of users who have not had a purchase history of product A and refers to a Web page related to science less than 20 times, and is related to health. It represents the occupancy rate of the number of users who do not have the purchase history of the product A with the user who has referred to the Web page less than once (that is, not referring to it).

  Similarly, the number “75.8” on the lower right in the frame of the node nd3 indicates that the number of web pages related to science in the total number of users who have the purchase history of the product A is 20 It represents the occupancy rate of the number of users who have a purchase history of the product A with users who are less than times and who have referred to health related web pages less than once (that is, not referring to them).

  Next, “number: 15310: 240” is displayed in the upper part of the frame of the node nd 4 which is another node of the hierarchy one level lower than the node nd 1. That is, the upper number in the frame of the node nd4 indicates the user included in the node nd4 (that is, among the analysis target users, the number of times the science related web page has been referenced less than 20 times and related to health Among users (whose number of times the web page is referred to is one or more), 15310 indicates that there is no purchase history of product A, and 240 indicates that there is a purchase history of product A. As can be understood from these figures, the number of users included in the node nd3 is a total value of 15,310 and 240, which is 15,550.

  In the middle part of the frame of the node nd4, “ratio: 98.5: 1.5” is displayed. That is, the middle numbers in the frame of the node nd4 indicate that among the analysis target users, the number of times of referring to the Web page related to science is less than 20 and the number of times of reference to the Web page related to health is 1 Among the users who are more than times, 98.5% of the users have no purchase history of the product A, and 1.5% of the users have a purchase history of the product A.

  In addition, in the lower part of the frame of the node nd4, “Oishi: 15.5: 24.0” is displayed. That is, the number on the left side of “15.5” is less than 20 times the number of times of reference to the science related web page in the total number of users without a purchase history of product A, and is related to health It represents the occupancy rate of the number of users who do not have a purchase history of the product A in a user whose number of times of reference to the Web page is one or more.

  Similarly, “24.0”, which is the lower right number in the frame of the node nd4, indicates that the number of times the science related Web page has been referred to in the total number of users who have a purchase history of the product A is 20. It represents the occupancy rate of the number of users who have a purchase history of the product A with a user who is less than the number of times and the number of times the web page related to health is referenced is one or more.

  The contents of the nodes nd5, nd6, nd7,..., Nd12, which are other nodes, are also the same as the contents of the nodes nd3 and nd4 described above. The description is omitted because it is similar.

As illustrated in FIG. 4, rules (branches) connecting between nodes included in the decision tree dt1 are displayed in different thicknesses. The decision tree dt1 in the present embodiment visually expresses the number of users included in each node according to the difference in rule thickness.
For example, the rules are displayed in a thickness proportional to the number of users included in the lower nodes to which the rules connect. For example, a rule eg1 connecting the node nd0 and the node nd1 is expressed by a thickness corresponding to “99990 people” which is the number of users included in the node nd1. Similarly, for example, a rule eg2 connecting the node nd0 to the node nd2 is expressed by a thickness corresponding to “10 people” which is the number of users included in the node nd2. Therefore, the thickness of the rule eg1 is displayed thicker than the thickness of the rule eg2.

  Also, as illustrated in FIG. 4, the colors in the frame of the node are displayed in different densities depending on the node. For example, the frames in the nodes nd6 and nd10 are displayed in dark color, and the nodes nd2 and nd12 are displayed in slightly dark color. The decision tree dt1 in the present embodiment visually expresses the magnitude of the value of the representative index described above by the color density in the frame of each node.

For example, as described above, the representative index in the present embodiment is the value of “ratio” itself, and in the decision tree dt1 shown in FIG. Is displayed in dark color. In addition, for a node whose value of the representative index is 20 or more, the frame of the node is displayed in a slightly dark color. In addition, for nodes whose representative index value is less than 20, the frame of the node is displayed in white.
The display method storage unit 20 shown in FIG. 2 stores display method information indicating a method of displaying nodes according to the value of the representative index. The display method information is, for example, two-dimensional tabular data including a series of two items of an item indicating the value of the representative index and an item indicating the display method of the node. The display method storage unit 20 is configured to include, for example, a storage medium such as a hard disk drive, a flash memory, an Eyprom, a ROM, a RAM, or any combination thereof.

  Here, the value of “ratio” is the value on the right side of the column of the ratio of each node, that is, the ratio of the user who is purchasing the product A in the user included in each node. For example, in the middle part of the frame of the node nd2, "ratio: 80.0: 20.0" is displayed. The right number in this display (that is, the value of the representative index) is “20.0”, which is a value of 20 or more, so the frame of the node nd2 is displayed in a slightly dark color. Also, for example, in the middle part of the frame of the node nd6, “ratio: 0.0: 100.0” is displayed. The right number in this display (that is, the value of the representative index) is “100.0” and has a value of 80 or more, so the frame of the node nd2 is displayed in dark color.

In each node of the decision tree dt1, if the number of users who have purchased the product A (that is, the number displayed on the right side of the upper row in the frame of each node) is less than the above-mentioned threshold value Even if the value of “ratio” is high, the color in the frame of the node is not displayed in dark or slightly dark (displayed in white).
For example, the decision tree dt1 illustrated in FIG. 4 is a decision tree when the value of the threshold value is “1”. In each of the nodes nd2, nd6, nd10, and nd12 in which the frame is displayed in dark or light color, the number of users who have purchased the product A at each node (ie, the frame of each node Since the number displayed on the right side of the upper row of the inside is larger than 1, the frame is not displayed in white. If the threshold value is "3", the frames of the nodes nd2, nd6, and nd12 whose number of users who have purchased the product A is 3 or less are displayed in white. become.

As described above, by visually displaying the number of users included in each node according to the thickness of each rule, a node having a large number of users can be easily recognized by the analyst.
Further, by displaying and emphasizing the color density in the frame of the node having a large value of the representative index high, the node having a large value of the representative index can be easily recognized by the analyst. Thereby, the analyst can easily recognize the node including the user group whose value of “ratio” which is the representative index is high, that is, the purchase rate of the product A is high.

  Furthermore, the color in the frame of the node, which is the number of users below the threshold set based on the number of users included in the node selected by the expert in the past, is not displayed with high density, The node is less likely to be recognized by the analyst (or is more likely to be recognized by the analyst as a node which is less likely to be selected by the expert).

Another functional configuration of the information processing apparatus 1 will be described with reference to FIG. 2 again.
The decision tree dt1 is presented to the analyst by the output unit 18 outputting an image including the decision tree dt1. The analyst selects a node from the decision tree dt1 based on the image displayed on the output unit 18, and inputs data indicating the selected node through the input unit 11.

The input unit 11 outputs data indicating the node selected by the analyst to the advertisement distribution unit 19.
The advertisement distribution unit 19 outputs data indicating a group of users included in a node based on the data input from the input unit 11 to the Web advertisement distributor terminal 3.

Further, the input unit 11 outputs, to the control unit 10, the node selected by the analyst and data indicating the skill level of the analyst. Alternatively, when the analyst who has input the data indicating the selected node is a skilled person, the input unit 11 may output data indicating the node selected by the analyst to the control unit 10.
It is assumed that the information processing apparatus 1 can recognize the skill level of the analyst in advance. For example, a table (not shown) indicating the skill level for each analyst is stored in advance in the information processing apparatus 1, and information is obtained by authenticating when the analyst starts using the information processing apparatus 1. The processing device 1 may be able to recognize the skill level of the analyst. Thereby, the information processing apparatus 1 can recognize the skill level of the analyst (for example, whether the analyst is a skilled person or not).

  The control unit 10 acquires the value of “number” corresponding to the node based on the data input from the input unit 11 from the decision tree calculation unit 15 and stores the value in the cutaway value storage unit 17 via the representative index calculation unit 16. Let Thereby, the information processing apparatus 1 can accumulate the number of users included in the node selected by the expert. The information processing device 1 can learn the value of the above-mentioned threshold value based on the accumulated data.

  The threshold value may be, for example, the minimum value among the values of the "number" stored in the threshold value storage unit 17, or the "number" stored in the threshold value storage unit 17. It may be an average value of the values of.

  In addition, when a plurality of nodes are selected by a skilled analyst in the past analysis, the value of “number” stored in threshold value storage unit 17 is each included in the plurality of selected nodes. It may be the minimum value among the number of users, or it may be the average value or the total value of the number of users.

(Operation of information processing apparatus)
Hereinafter, the operation of the information processing apparatus 1 will be described with reference to the drawings.
FIG. 5 is a flowchart showing the operation of the information processing apparatus 1 according to the present embodiment.
The process shown in the present flowchart is started when data indicating an analysis request from an analyst is input to the input unit 11 of the information processing device 1.

(Step S101) The input unit 11 receives an analysis request from an analyst. The input unit 11 outputs, to the control unit 10, data indicating an analysis request, which is input from the analyst. Thereafter, the process proceeds to step S102.

(Step S102) When data indicating an analysis request from the analyst is input from the input unit 11, the control unit 10 outputs the data to the simulation verification evaluation unit 14. The simulation verification evaluation unit 14 acquires analysis target data from the history storage unit 12. The analysis target data includes user attribute information for each user, for example, the purchase history of the product A, the browsed Web page, and the category to which the content of the browsed Web page belongs (that is, data indicating the preference of the user). The simulation verification evaluation unit 14 extracts analysis target data in a plurality of different periods from the analysis target data acquired from the history storage unit 12 and outputs the analysis target data to the decision tree calculation unit 15 via the control unit 10, respectively. Thereafter, the process proceeds to step S103.

(Step S103) The decision tree calculation unit 15 calculates a decision tree based on analysis target data in a plurality of different periods input from the simulation verification and evaluation unit 14 via the control unit 10. The decision tree calculation unit 15 outputs data indicating each calculated decision tree to the simulation verification evaluation unit 14 via the control unit 10. The simulation verification and evaluation unit 14 outputs data indicating a decision tree based on analysis target data in a plurality of different periods to the representative index calculation unit 16 via the control unit 10. Thereafter, the process proceeds to step S104.

(Step S104) The representative index calculation unit 16 calculates a representative index for each node of the decision tree based on the data input from the control unit 10. Thereafter, the process proceeds to step S105.

(Step S105) The representative index calculation unit 16 controls the data indicating the value of the representative index of each node of the decision tree for each of the plurality of periods and the data indicating the decision tree for each of the plurality of periods via the control unit 10. Output to the simulation verification and evaluation unit 14.
The simulation verification and evaluation unit 14 selects a decision tree in which the feature of the value of the representative index appears most clearly, based on the value of the representative index of each node of the decision tree for each of a plurality of periods. The simulation verification and evaluation unit 14 outputs data indicating the selected decision tree to the representative index calculation unit 16 via the control unit 10. Thereafter, the process proceeds to step S106.

(Step S106) The representative index calculation unit 16 is the value of the “number” of nodes selected by the skilled analyst for the value of “number” in each node of the decision tree selected by the simulation verification evaluation unit 14 If the value is less than the "threshold value", the value of the representative index of the node is set to 0. That is, the value of the representative index of nodes whose number of users is less than the threshold value is set to 0. The representative index calculation unit 16 outputs data indicating the decision tree and data indicating the representative index of each node to the control unit 10. Thereafter, the process proceeds to step S107.

(Step S107) The control unit 10 generates data indicating an image including a decision tree, based on the data indicating the decision tree and the data indicating the representative index of each node, which are input from the representative index calculation unit 16. The control unit 10 outputs data indicating an image including the calculated decision tree to the output unit 18. Thereafter, the process proceeds to step S108.

(Step S108) The output unit 18 outputs an image including a decision tree based on the data input from the control unit 10. Thereby, an image including the decision tree is presented to the analyst. Thereafter, the process proceeds to step S109.

(Step S109) The analyst selects a node in the decision tree (that is, a node including a user group to be targeted for advertisement distribution is selected), and the input unit 11 receives an input of data indicating the selected node. In the case, the process proceeds to step S110. If not, the process remains in step S109.

(Step S110) The input unit 11 outputs data indicating the selected node to the advertisement distribution unit 19. The advertisement distribution unit 19 distributes the Web advertisement to the group of users included in the node based on the data input from the input unit 11 via the Web advertisement distribution provider. Thereafter, the process proceeds to step S111.

(Step S111) If the analyst who has input the data indicating the selected node is a skilled person, the input unit 11 proceeds to step S112. If not, the processing of this flowchart ends.

(Step S112) The input unit 11 outputs data indicating the node selected by the analyst to the control unit 10. The control unit 10 acquires the value of “number” corresponding to the node based on the data input from the input unit 11 from the decision tree calculation unit 15. The control unit 10 stores the value of “number” based on the data acquired from the decision tree calculation unit 15 in the cutoff value storage unit 17 via the representative index calculation unit 16. Thereby, the information processing device 1 accumulates the number of users included in the node selected by the expert, and the information processing device 1 learns the threshold value. This is the end of the processing of this flowchart.

As described above, the information processing apparatus 1 according to the present embodiment sets a plurality of analysis target periods and calculates a decision tree based on log data (user attribute information) based on each analysis target period, as described above. Do. The information processing device 1 calculates the value of the representative index of each node of the calculated decision tree. The information processing apparatus 1 selects a decision tree in which the feature of the value of the representative index is most clearly displayed, based on the value of the representative index of each decision tree. The information processing apparatus 1 selects a decision tree in which the characteristic of the value of the representative index is most clearly displayed and presents the decision tree to the analyst, thereby making the decision tree capable of further increasing the effect on the task of analysis. It can be presented to the analyst.
As described above, the information processing apparatus 1 according to the present embodiment can present the analysis result in which the feature is more strongly exhibited among the plurality of analysis results in which the analysis target period is different.

  The embodiment of the present invention has been described in detail, but the specific configuration is not limited to the above-described one, and various design changes can be made without departing from the scope of the present invention. is there.

  Note that part or all of the information processing apparatus 1 in the embodiment described above may be realized by a computer. In that case, a program for realizing the control function may be recorded in a computer readable recording medium, and the program recorded in the recording medium may be read and executed by a computer system.

  Here, the “computer system” is a computer system built in the information processing apparatus 1 and includes an OS and hardware such as peripheral devices. The term "computer-readable recording medium" refers to a storage medium such as a flexible disk, a magneto-optical disk, a ROM, a portable medium such as a ROM or a CD-ROM, or a hard disk built in a computer system.

  Furthermore, the “computer-readable recording medium” is one that holds a program dynamically for a short time, such as a communication line when transmitting a program via a network such as the Internet or a communication line such as a telephone line. In this case, a volatile memory in a computer system serving as a server or a client in that case may also include one that holds a program for a certain period of time. The program may be for realizing a part of the functions described above, or may be realized in combination with the program already recorded in the computer system.

  In addition, the information processing device 1 in the above-described embodiment may be realized as an integrated circuit such as an LSI (Large Scale Integration). Each functional block of the information processing apparatus 1 may be individually processorized, or part or all may be integrated and processorized. Further, the method of circuit integration is not limited to LSI's, and implementation using dedicated circuitry or general purpose processors is also possible. In the case where an integrated circuit technology comes out to replace LSI's as a result of the advancement of semiconductor technology, integrated circuits based on such technology may also be used.

1 ... information processing apparatus, 2 ... Web server, 3 ... Web advertisement distributor terminal, 10 ... control unit, 11 ... input unit, 12 ... history storage unit, 13 ... Site preference definition unit 14 Simulation verification evaluation unit 15 Decision tree calculation unit 16 Representative index calculation unit 17 Deadline value storage unit 18 Output unit 19 ... Advertising distribution unit, 20 ... Display method storage unit

Claims (6)

A decision tree for acquiring data indicating user attributes of each of a plurality of users, and classifying the users into any of a plurality of nodes based on the user attributes of data corresponding to a specific period of the data, A decision tree calculation unit that calculates a plurality of times while changing a period;
A representative index calculation unit that calculates a value of a representative index that is an index based on the number of users classified into the nodes;
A verification and evaluation unit that selects any decision tree from among a plurality of decision trees calculated by the decision tree calculation unit based on the value of the representative index calculated by the representative index calculation unit;
An output unit that outputs analysis result information including a decision tree calculated based on data indicating the decision tree selected by the verification and evaluation unit and data indicating the value of the representative index calculated by the representative index calculation unit;
An information processing apparatus comprising: The verification and evaluation unit calculates the variance of the values of the representative index for each of the nodes calculated by the representative index calculation unit for each of the plurality of different period-based decision trees, and selects the decision tree with the largest variance. Select a decision tree more characterized by the value of the representative index by
An information processing apparatus according to claim 1, characterized in that. The representative index is a ratio of the number of users who acted as an analysis target to the number of users classified into the node.
The information processing apparatus according to claim 1 or 2, characterized in that: The representative index is a value obtained by multiplying the average number of sales per user of the users classified into the node by the ratio of the number of users who performed the analysis target out of the number of users classified into the nodes.
The information processing apparatus according to any one of claims 1 to 3, characterized in that: An information processing method using a computer,
The decision tree calculation unit included in the computer acquires data indicating user attributes of each of a plurality of users, and the user is selected from any of a plurality of nodes based on the user attributes of data corresponding to a specific period among the data. A decision tree calculating step of calculating a plurality of decision trees to be classified into different ones by changing the period;
A representative index calculation step of calculating a value of a representative index calculated based on the number of the users classified into the nodes, the representative index calculation unit included in the computer;
The verification evaluation unit included in the computer determines any one decision tree among the plurality of decision trees calculated in the decision tree calculation step based on the value of the representative index calculated in the representative index calculation step. The verification evaluation step to select
Analysis showing a decision tree calculated based on data indicating the decision tree selected in the verification and evaluation step and data indicating the value of the representative index calculated in the representative index calculating step, and an output unit included in the computer An output step for outputting result information;
An information processing method characterized by comprising: On the computer
A determination tree for acquiring data indicating user attributes of a plurality of users and classifying the users into any of a plurality of nodes based on the user attributes of data corresponding to a specific period among the data, the period Decision tree calculation step of calculating plural numbers by changing
A representative indicator calculation step of calculating the value of the representative indicator calculated based on the number of the users classified into the nodes;
A verification and evaluation step of selecting any decision tree from among a plurality of decision trees calculated in the decision tree calculating step, based on the value of the representative index calculated in the representative index calculating step;
An output step of outputting analysis result information indicating a decision tree calculated based on data indicating the decision tree selected in the verification and evaluation step and data indicating the value of the representative index calculated in the representative index calculating step; ,
A program to run a program.

Images