JP2015166989A - information processing apparatus and information analysis method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To improve the accuracy of information to be obtained by analyzing customer information.SOLUTION: An information analysis device 10 includes a customer information holding section 20, an area information holding section 22, and a model generating section 32. The customer information holding section 20 stores customer information on a customer. The area information holding section 22 stores resident attributes for a plurality of areas. The model generating section 32 acquires customer information on a first customer and a second customer stored in the customer information holding section 20, and the resident attributes for the areas where the first customer and the second customer live, stored in the area information holding section 22, on the first and second customers having different purchase probabilities for a commodity to be analyzed. The model generating section 32 generates a mathematical model by using the purchase probability for the commodity to be analyzed as an explained variable, and both customer information and resident attribute as an explanatory variable.

Description

  The present invention relates to data processing technology, and more particularly to information processing technology.

  In recent years, attempts have been made to obtain useful information in business by causing a computer to analyze various information. In the following Patent Document 1, the present applicant proposes a technique for analyzing product sales information and creating information for supporting user sales activities. Specifically, with regard to the region where the purchaser of the product resides (hereinafter also referred to as “area”), paying attention to the area cluster that is a group of regions where the attributes of the residents are similar, based on the area cluster where the purchaser resides The company proposes a technology that creates useful information for merchandise sellers by analyzing sales information.

JP 2009-169698 A

  Many companies manage customer information in-house, and some companies analyze customer information to obtain useful information for corporate activities. Such a technique is also called data mining. The present inventor recognizes that the accuracy of data mining using only customer information accumulated in the company is not necessarily high, and further increases the accuracy of data mining by further adding data of the area where the customer resides. I thought it could be improved.

  The present invention has been made based on the above-mentioned recognition of the present inventor, and its main purpose is to provide a technique for improving the accuracy of information obtained by analyzing customer information.

  In order to solve the above-described problems, an information processing apparatus according to an aspect of the present invention includes a first storage unit that stores customer information related to an individual customer, a second storage unit that stores inhabitant attributes of a plurality of areas, and an analysis target. For the first customer and the second customer having different purchase probabilities for the product, customer information of the first customer and the second customer stored in the first storage unit, and the first customer stored in the second storage unit And a model generation unit that generates a mathematical model having the purchase probability for the product as an explained variable and both the customer information and the resident attribute as explanatory variables based on the resident attribute of the area where the second customer resides.

  Another aspect of the present invention is an information analysis method. The method includes a first step of acquiring customer information relating to a first customer individual having a first purchase accuracy for a product to be analyzed, a resident attribute of an area where the first customer resides, and a first step for the product. Acquired by the second step, the first step, and the second step of acquiring customer information regarding the second customer individual having the second purchase accuracy different from the purchase accuracy of 1, and the resident attribute of the area where the second customer resides Based on the obtained data, the computer executes a step of generating a mathematical model having the purchase accuracy for the product as the explained variable and both the customer information and the resident attribute as the explanatory variables.

  It should be noted that any combination of the above-described constituent elements and a representation of the present invention converted between a system, a program, a recording medium storing the program, etc. are also effective as an aspect of the present invention.

  According to the present invention, it is possible to support improvement in accuracy of information obtained by analyzing customer information.

It is a block diagram which shows the function structure of the information analyzer of embodiment. It is a figure which shows the example of the customer information hold | maintained at a customer information holding part. It is a figure which shows the example of the resident attribute record hold | maintained at an area information holding part. It is a figure which shows the list information of the purchase probability of each customer.

  Until now, many companies have been trying to obtain information useful for business by performing data analysis work (for example, data mining by computer) using only customer information managed by the company. For example, in the case of a bank, a customer to whom a direct mail is to be sent is selected by analyzing internal data such as a customer's name, age, and own deposit amount stored in the bank's database. However, the internal data is basically limited to data related to transaction results with customers and data disclosed by customers, and the accuracy of analysis results using only internal data is not necessarily high.

  Therefore, the information analysis apparatus according to the embodiment uses profile data such as estimated customer financial assets and financial needs estimated from customer address data that has been managed as customer information but has not been fully utilized. And build a mathematical model (recommendation model) that calculates the degree of product recommendation for each customer. In other words, in addition to the company's internal data, a mathematical model that reflects external data indicating the customer's residence characteristics is constructed. The mathematical model of the embodiment is a mathematical formula obtained as a result of multivariate analysis, for example, a regression formula. In the embodiment, a mathematical model (hereinafter referred to as a “purchase accuracy model”) for obtaining the purchase accuracy of each customer for a product sold by a company is constructed.

  For example, a specific customer residing in an area has an internal deposit of millions of yen as internal data, but the estimated deposits across multiple financial institutions indicated by the characteristics of residence as external data is 80 million yen. It may be. In this case, the information analysis apparatus according to the embodiment constructs a purchase probability model (the target product is an investment trust) in consideration of the estimated deposit amount of the customer. Further, from the constructed purchase accuracy model, the customer's purchase accuracy is obtained by including the estimated deposit amount of the customer in the parameter.

  As another example, a particular customer residing in an area currently has only car insurance, but the customer's residence characteristics may indicate a very high percentage of households with children. In this case, the information analysis apparatus according to the embodiment constructs a purchase probability model (the target product is an educational insurance) in consideration of the proportion of households with children. Further, from the constructed purchase accuracy model, the purchase probability of the customer is obtained by including, as a parameter, the percentage of households with children in the residence of the customer. Thereby, the information accuracy (accuracy) estimated by the purchase accuracy model is improved, and the sales promotion of the product is supported. It should be noted that the “purchase accuracy” in the embodiment can be said to be a purchase possibility and a response rate to the sales promotion action. In addition, the “product” in the embodiment is a concept including a right and a service sold by a company.

  FIG. 1 is a block diagram illustrating a functional configuration of the information analysis apparatus according to the embodiment. The information analysis apparatus 10 is an information processing apparatus such as a PC installed in a company. The information analysis device 10 includes a control unit 12 and a storage unit 14. The control unit 12 executes various data processing, and particularly executes information analysis processing. The storage unit 14 is a storage area that stores data that is referred to and updated by the control unit 12. The storage unit 14 also provides a work area for the control unit 12.

  Each block shown in the block diagram of the present specification can be realized in terms of hardware by an element such as a CPU of a computer or a mechanical device, and in terms of software, it can be realized by a computer program or the like. The functional block realized by those cooperation is drawn. Therefore, those skilled in the art will understand that these functional blocks can be realized in various forms by a combination of hardware and software.

  For example, an information analysis application including a program module corresponding to each functional block in FIG. 1 may be stored in a predetermined recording medium and installed in the storage of the information analysis apparatus 10. When the user activates an information analysis application in the information analysis apparatus 10, the CPU of the information analysis apparatus 10 executes these program modules while reading them into the main memory as appropriate, so that the function of each functional block is exhibited. Good. 1 may be executed in a distributed environment by cooperation of a plurality of devices (for example, a web client terminal such as a PC equipped with a web browser, a web application server, a database server, etc.) via a communication network. Good.

  The storage unit 14 includes a customer information holding unit 20, an area information holding unit 22, and a model holding unit 24. The model holding unit 24 is a purchase accuracy model generated by a model generation unit 32 described later, and holds data indicating models of one or more products to be analyzed.

  The customer information holding unit 20 holds customer information managed by the company as customer information. Customers include not only individuals and organizations that have purchased at least some of the products and services sold by companies, but also individuals and organizations that have not purchased products and services. The latter is, for example, a consumer who answered a questionnaire or survey conducted by a company and applied for a present. The customer information typically includes information indicating transaction results between the company and the customer, and includes customer attribute information and so-called personal information disclosed by the customer to the company.

  FIG. 2 shows an example of customer information held in the customer information holding unit 20. In the embodiment, an example will be described in which a company that is an information analysis entity is an insurance company, and a mathematical model for calculating the purchase accuracy of fire insurance and earthquake insurance for each customer is constructed. The customer information record of each customer held by the customer information holding unit 20 includes at least the address where the customer resides (housing display, lot number, etc.) and information indicating whether the customer has already purchased the product to be analyzed. For example, the fire insurance payment amount column in FIG. 2 indicates the amount of fire insurance premium paid to the company. If it is greater than 0, fire insurance has been contracted (purchased), and if 0, fire insurance has not been contracted ( Unpurchased). The same applies to the earthquake insurance payment amount column in FIG.

  Returning to FIG. 1, the area information holding unit 22 includes information indicating attributes relating to residents living in each area (hereinafter “residents”) for each of a plurality of geographically partitioned areas, in other words, a plurality of areas on the map. Also called “attribute record”. Specifically, the area information holding unit 22 sets each area of about 180,000 town-chomes in Japan as an area, and holds a resident attribute record of each area. In the area position column, the location information at the town level is stored.

  Resident attributes of each area include demographic attributes indicating demographic attributes for each area, and psychographic attributes indicating attributes related to human psychology such as values and lifestyles of the residents of each area. Examples of demographic attributes include “nuclear family with relatively small children in their 30s and 40s”, “income is slightly higher than average, and many are university graduates”, “two children”, etc. . On the other hand, as an example of psychographic attributes, “For women in their 20s, brand, safety and economy are very important, but there is little environmental orientation”, “For women in their 30s, the brand is slightly emphasized, “Environmentally oriented” and “relatively high risk tolerance”. In other words, the resident attribute of each area is information indicating a consumer image living in each area, and includes the age, income level, occupation, educational background, family structure, living environment, preferences, way of thinking, and the like of the consumer.

  Specifically, an index value indicating the resident attribute of each area is recorded in the resident attribute record held by the area information holding unit 22. This index value is a plurality of types of index values obtained by indexing the degree to which the resident attribute matches each of a plurality of types of evaluation criteria with a predetermined evaluation function. The index value includes, for example, an average annual income, an average number of households, an average number of children per household, a 30s ratio, a 40s ratio, and the like. In other words, it can be said that the resident attribute record held by the area information holding unit 22 records qualitative resident attributes converted into a plurality of types of numerical values for calculation processing by a computer. In addition, if it is originally a quantitative resident attribute, it may be recorded in the resident attribute record without conversion.

  FIG. 3 shows an example of a resident attribute record held in the area information holding unit 22. Inhabitants attribute records are (1) various publicly available statistical information such as national census, (2) original estimation information by developers of information analysis applications, (3) questionnaires conducted for residents in each area Contains result information. The statistical information of (1) includes age ratio, gender ratio, occupation ratio, education ratio, etc., and “owned house ratio” and “average number of children” in FIG. The estimation information in (2) includes average income, average assets, average land prices, etc., and “income level” in FIG. The result information of the questionnaire in (3) includes response results such as lifestyle, values, and consumption preferences, and corresponds to “safety orientation”, “fire insurance potential demand”, and “earthquake insurance potential demand” in FIG. . As described above, qualitative information such as the income level and safety orientation shown in FIG. 3 is converted into an index value in advance by a predetermined evaluation function. Each item of area information, particularly an item obtained by converting qualitative information into an index value, may be adjusted as appropriate according to the user's judgment in the process of information analysis.

  Returning to FIG. 1, the control unit 12 includes an operation detection unit 30, a model generation unit 32, a purchase accuracy calculation unit 36, and a customer extraction unit 38. The operation detection unit 30 detects a user operation input to an input device such as a keyboard or a mouse, and notifies the user input content to other functional blocks. Each functional block of the information analysis device 10 receives user input information via the operation detection unit 30.

  Hereinafter, a plurality of customers having different purchase probabilities for the analysis target product designated by the user will be referred to as “first customer” and “second customer” for convenience. The model generation unit 32 includes the customer information of the first customer and the second customer held in the customer information holding unit 20, and the resident attribute of the area where the first customer and the second customer held in the area information holding unit 22 reside. Based on both, a purchase accuracy model for obtaining the purchase accuracy of each customer for the analysis target product is generated. The model generation unit 32 stores data indicating the generated purchase accuracy model in the model holding unit 24 in association with identification information of an analysis target product (product indicating purchase accuracy) based on the model.

  The model generation unit 32 identifies a customer who has already purchased the analysis target product as the first customer, and identifies an unpurchased customer as the second customer. For example, when the analysis target product is fire insurance, a customer storing a value of 1 or more in the fire insurance payment amount column in the table of FIG. 2 is identified as a first customer, and a customer storing 0 is identified as a second customer. . When generating the purchase accuracy model, the model generation unit 32 reads both the customer information record of the first customer and the customer information record of the second customer from the customer information holding unit 20. The model generation unit 32 searches the area information holding unit 22 for the area to which the address recorded in the customer information of each customer belongs. Specifically, the model generation unit 32 identifies and reads a record in which an area position that matches the customer address in front of the customer attribute record of each area is recorded as a customer attribute record of the customer's residential area.

  In other words, the model generation unit 32 generates both the customer information relating to the first customer having the first purchase accuracy with respect to the analysis target product and the inhabitant attribute of the residence area of the first customer. Get as a parameter for. Further, the model generation unit 32 has both customer information regarding the second customer having a second purchase accuracy different from the first purchase accuracy for the analysis target product and the resident attribute of the second customer's residence area. Are obtained as parameters for generating a purchase accuracy model.

  The model generation unit 32 includes an analysis engine 34. The analysis engine 34 executes multivariate analysis processing based on a known method such as multiple regression analysis or logistic regression analysis. The analysis engine 34 may be known multivariate analysis software. The model generation unit 32 may call an API provided by the analysis engine 34 and acquire data of a mathematical model constructed as a result of the multivariate analysis process from the analysis engine 34.

  The model generation unit 32 inputs the purchase accuracy of each of the plurality of customers, the customer information of each customer, the resident attributes of the residential area, and the identification information of the multivariate analysis method designated by the user to the analysis engine 34. In the embodiment, the purchase accuracy of the first customer is designated as “1”, and the purchase accuracy of the second customer is designated as “0”. Instead of such a discrete value, a continuous value such as a payment amount may be specified as the purchase accuracy. In addition, a specific purchase accuracy may be selected and specified from a plurality of predetermined step values according to the customer's payment amount or the like.

The analysis engine 34 executes a multivariate analysis process according to the multivariate analysis method designated by the user, generates a purchase accuracy model, and outputs it to the model generation unit 32. The analysis engine 34 generates a mathematical model having both the customer information item and the resident attribute item as explanatory variables as the purchase accuracy model, and the purchase accuracy for the analysis target product as the explained variable (object variable). An example of a purchase accuracy model, here, a regression equation indicating the purchase accuracy model is shown below.

  X1 and X2 in Formula 1 indicate items of customer information as internal data. For example, it may be age, transaction results, and existing contract information. In addition, Y1 and Y2 in Formula 1 indicate items of residential area inhabitant attributes as external data. For example, the ownership ratio, income level, potential demand for fire insurance, and potential demand for earthquake insurance may be used. In Equation 1, a, b, c, and d are values indicating the degree of influence of each information item on the purchase accuracy, and can be said to be regression coefficients. The analysis engine 34 calculates the values of the coefficients a, b, c, and d by multivariate analysis processing.

  The purchase accuracy calculation unit 36, with respect to the purchase accuracy model generated by the model generation unit 32, the analysis target product (product for which purchase accuracy is obtained) in the model, customer information of the unpurchased customer, and the residence of the customer By inputting the resident attribute of the area, the purchase accuracy of the analysis target product by the customer is calculated. The purchase accuracy calculation unit 36 stores purchase accuracy data for one or more unpurchased customers in a predetermined area of the storage unit 14 in association with the ID and name of the customer.

  Specifically, the purchase accuracy calculation unit 36 reads a purchase accuracy model (a purchase accuracy model related to a specific product) designated by the user among the purchase accuracy models stored in the model holding unit 24. Further, the purchase accuracy calculation unit 36 is a customer who has not purchased the product targeted by the purchase accuracy model specified by the user in the customer information held by the customer information holding unit 20, for example, the transaction record is not recorded. The customer or the customer whose payment amount is 0 is specified, and the customer information of the customer and the resident attributes of the living area are read. The purchase accuracy calculation unit 36 inputs the customer information of the unpurchased customer for the target product of the model and the resident attribute of the residence area for the purchase accuracy model specified by the user, thereby obtaining the purchase accuracy of the customer. calculate.

  The customer extraction unit 38, for a customer whose purchase accuracy calculated by the purchase accuracy calculation unit 36 is a predetermined threshold or more among a plurality of customers who have not purchased the target product of the purchase accuracy model, the sales promotion action for the target product is effective. Extract as customers who are easy to act on. This sales promotion action includes, for example, sending of direct mail, telephone solicitation, and door-to-door sales.

  Specifically, the customer extraction unit 38 refers to the purchase accuracy data of one or more unpurchased customers stored in a predetermined storage area by the purchase accuracy calculation unit 36 based on the purchase accuracy model specified by the user. Information indicating that a customer whose purchase accuracy is equal to or higher than a predetermined threshold is a target customer who should carry out a sales promotion action for the analysis target product and is a so-called prospective customer with a relatively high possibility of purchasing the product. Are stored in a predetermined storage area. An appropriate value may be set as the threshold value of the purchase accuracy for specifying the prospective customer based on the knowledge of the user or the information analysis application developer, the experiment using the information analysis device 10, or the like.

The operation of the information analysis apparatus 10 having the above configuration will be described below.
A user in charge of customer information analysis for product sales promotion constructs an environment so that customer information and area information can be accessed from the information analysis apparatus 10. For example, the customer information holding unit 20 and the area information holding unit 22 are constructed in the storage unit 14 of the information analysis apparatus 10. As a modification, the information analysis apparatus 10 may be set to be accessible to the customer information database and the area information database. The user inputs a sales accuracy model generation instruction to the information analysis apparatus 10. At this time, the user also designates a sales promotion target product (analysis target product by the information analysis apparatus 10) and an analysis method (types such as multiple regression analysis and logistic regression analysis).

  The model generation unit 32 acquires customer information managed by the company regarding the individual customer and the resident attributes of the customer's residence area for each of the customers who have purchased the analysis target product and the unpurchased customers. Then, using a known multivariate analysis method, a purchase accuracy model is generated that indicates the degree of influence each of the customer information item and the residential area inhabitant attribute item has on the purchase accuracy of the analysis target product.

  The user may change the analysis method as appropriate, and may select items to be used for analysis among the items of customer information and resident attributes as appropriate, and generate multiple purchase accuracy models for one analysis target product. May be. Then, it may be determined which is most appropriate among a plurality of models. The user inputs a product recommendation degree estimation instruction to each customer into the information analysis apparatus 10 while designating a specific purchase accuracy model.

  The purchase accuracy calculation unit 36 inputs the customer information of each customer and the resident attribute of the residence area for a plurality of customers who have not purchased the target product of the model with respect to the purchase accuracy model specified by the user. The purchase accuracy of the target product by the customer is calculated, and list information of the purchase accuracy of each customer is generated. The customer extraction unit 38 extracts a customer (in other words, a prospective customer) whose purchase accuracy of the target product is relatively high among a plurality of customers who have not purchased the target product. The customer extraction unit 38 may add information for distinguishing the prospective customer from other customers (customers whose purchase accuracy of the target product is assumed to be low) to the purchase accuracy list information of each customer. Alternatively, only the information of prospective customers may be extracted from the list of purchase accuracy of each customer, and the prospective customer information may be recorded.

  FIG. 4 shows the list information of purchase accuracy of each customer. The recommended degree of fire insurance in the figure is the output data of the purchase probability model for the product “fire insurance”, and indicates the purchase accuracy of each customer for the product “fire insurance”. Further, the recommended degree of earthquake insurance in the figure is output data of the purchase accuracy model for the product “earthquake insurance”, and indicates the purchase accuracy of each customer for the product “earthquake insurance”.

  In the example of FIG. 4, the purchase accuracy calculation unit 36 calculates the purchase accuracy for the users who have purchased each insurance (contracted), and refers to the customer information holding unit 20 to determine whether or not the purchase has been completed (contracted). Is added to the list information. The customer extraction unit 38 determines that a customer whose contract status is not contracted and whose recommendation degree is 50% or more is a prospective customer, sets the character size of the recommendation degree of the prospective customer larger than other customers, A check mark indicating a prospective customer is added to the list information.

  As described above, according to the information analysis apparatus 10 of the embodiment, in addition to customer information as internal data individually managed by a company, a resident attribute of a customer's residence area is used to create a highly accurate purchase accuracy model. Can be provided. As a result, it is possible to support acquisition of more useful information in business and highly accurate information beyond the limits of internal data analysis. In addition, for customers who have not purchased the target product of the purchase accuracy model, the purchase accuracy of the customer is presented to the analyst, and the prospective customer is also presented to support efficient implementation of product promotion actions. it can.

  The present invention has been described based on the embodiments. This embodiment is an exemplification, and it will be understood by those skilled in the art that various modifications can be made to combinations of the respective constituent elements and processing processes, and such modifications are also within the scope of the present invention. is there.

  It should also be understood by those skilled in the art that the functions to be fulfilled by the constituent elements described in the claims are realized by the individual constituent elements shown in the embodiments and the modification examples or by their cooperation.

  DESCRIPTION OF SYMBOLS 10 Information analyzer, 20 Customer information holding part, 22 Area information holding part, 32 Model production | generation part, 36 Purchase accuracy calculation part, 38 Customer extraction part.

Claims (4)

A first storage unit for storing customer information relating to individual customers;
A second storage unit for storing inhabitants attributes of a plurality of areas;
For the first customer and the second customer having different purchase probabilities for the product to be analyzed, the customer information of the first customer and the second customer stored in the first storage unit, and stored in the second storage unit A model generation unit that generates a mathematical model having the purchase accuracy for the product as an explained variable and both the customer information and the resident attribute as explanatory variables based on the resident attribute of the area where the first customer and the second customer reside. When,
An information processing apparatus comprising:   The customer's purchase information of the customer who has not purchased the product and the resident attribute of the area where the customer resides are input to the mathematical model generated by the model generation unit, thereby calculating the purchase accuracy of the product by the customer. The information processing apparatus according to claim 1, further comprising a purchase accuracy calculation unit.   A customer extraction unit that extracts a customer whose purchase accuracy calculated by the purchase accuracy calculation unit is equal to or greater than a predetermined threshold among a plurality of customers who have not purchased the product as a customer whose sales promotion action for the product is effective; The information processing apparatus according to claim 2, further comprising: A first step of acquiring customer information relating to a first customer individual having a first purchase accuracy with respect to a product to be analyzed; and a resident attribute of an area where the first customer resides;
A second step of obtaining customer information relating to a second customer individual having a second purchase accuracy different from the first purchase accuracy for the product, and a resident attribute of an area where the second customer resides;
Based on the data acquired by the first step and the second step, generating a mathematical model with the purchase accuracy for the product as an explained variable and both customer information and resident attributes as explanatory variables;
A computer-executable information analysis method.

Images