JP2005135167A - Device, method, and program for data analysis - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a device, a method, and a program for data analysis that accurately and easily provide marketing data meeting customer's needs through data analysis wherein quantitative data and qualitative data are merged. <P>SOLUTION: The data analyzing device equipped with a database stored with quantitative data linked to the identifier of an object to be analyzed and a database stored with qualitative data linked to the identifier of the object to be analyzed integrates the quantitative data and qualitative data by linking them with the identifier of the object to be analyzed, inputs retrieval conditions for quantitative data, and outputs a retrieval result consisting of quantitative data meeting the retrieval conditions and qualitative data linked with the quantitative data meeting the retrieval conditions. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

  The present invention relates to a data analysis apparatus, a data analysis method, and a data analysis program for executing data analysis in which quantitative data and qualitative data are fused.

  In recent years, consumers have a variety of lifestyles and values, to accurately grasp the needs of consumers who are changing rapidly, and to develop products that can sell and improve customer satisfaction. It is very important to listen to "and listen to the real intentions hidden behind words." Due to the recent spread of Internet research call centers, etc., there is an environment that can collect and accumulate a large amount of “live voices”, but it takes a lot of time and labor to analyze, so the data is sufficient It is not utilized.

In order to solve such a problem, a marketing system as disclosed in Patent Document 1 has been conventionally considered. According to the present invention, a marketing system is constructed by connecting a plurality of monitor-side information processing terminals, a customer-side information processing terminal, and a server provided in the center via a network. Then, the server creates request information and selects a monitor according to the request summary instruction received from the customer, receives and stores response information for the request information from the selected monitor, and based on the analysis request instruction received from the customer, An analysis result is created from the stored answer information. When the created analysis result is transmitted to the customer, the monitor receives the request information and inputs the reply information. On the other hand, the customer transmits a request summary instruction and an analysis request instruction, and receives and outputs the analysis result transmitted from the server. By configuring in this way, it is possible to provide marketing data that meets customer needs accurately and easily.
JP 2002-133065 A

  According to the above-described marketing system, when viewed from a certain aspect, marketing data that meets customer needs is extracted. However, this is limited to the analysis of consumer purchase history data by data mining, etc., and it is not always possible to meet the need for marketing data that is effective in developing more sellable products and improving customer satisfaction. There was no problem.

  The present invention has been made in consideration of such circumstances, and its purpose is to provide accurate and easy marketing data that meets customer needs through data analysis that combines quantitative data and qualitative data. The present invention provides a data analysis apparatus, a data analysis method, and a data analysis program.

  The present invention has been made to solve the above-described problems. The present invention relates to first storage means for storing quantitative data associated with an identifier to be analyzed, and associated with the identifier to be analyzed. The second storage means for storing qualitative data, the data integration means for linking the quantitative data and the qualitative data with the identifier of the analysis target, the search condition for the quantitative data is input, and the search condition is met And data search means for outputting a search result including quantitative data corresponding to the search condition and qualitative data associated with the search condition by the data integration means.

  The present invention further includes text mining means for classifying the qualitative data into one or a plurality of classification hierarchies and adding a predetermined classification code, wherein the data search means includes the quantitative data search condition and the text mining The classification code added by the means is input, and the quantitative data and the qualitative data corresponding to the quantitative data search condition and linked by the data integration means including the classification code are output as search results. It is characterized by doing.

  Further, according to the present invention, the text mining means classifies the qualitative data into a plurality of different hierarchies from a large classification to a small classification, adds a predetermined classification code, and the data search means The data search condition and one or more classification codes from the large classification to the small classification added by the text mining means are input, and the classification code is matched with the quantitative data retrieval condition. The quantitative data and the qualitative data linked by the data integration means are output as search results.

  In addition, the present invention includes first storage means for storing quantitative data associated with an identifier to be analyzed, and second storage means for storing qualitative data associated with the identifier to be analyzed. A data analysis method in the data analysis apparatus for performing the analysis by linking and integrating the quantitative data and the qualitative data with the identifier of the analysis target, inputting a search condition for the quantitative data, and corresponding to the search condition A search result comprising quantitative data and qualitative data associated with the quantitative data corresponding to the search condition is output.

  In addition, the present invention includes first storage means for storing quantitative data associated with an identifier to be analyzed, and second storage means for storing qualitative data associated with the identifier to be analyzed. A program for causing a data analysis apparatus to execute data analysis processing, wherein the quantitative data and the qualitative data are linked by the identifier of the analysis target and integrated, and a search condition for the quantitative data is input. A data analysis program for executing a process for outputting a search result including quantitative data corresponding to the search condition and qualitative data associated with the quantitative data corresponding to the search condition.

As described above, according to the present invention, the quantitative data and the qualitative data are linked and integrated by the identifier of the analysis target, the search condition for the quantitative data is input, the quantitative data corresponding to the search condition, A search result consisting of quantitative data corresponding to the search condition and qualitative data associated with the search condition is output.
With this configuration, a large amount of analysis objects are narrowed down by a search condition regarding desired quantitative data, and qualitative data associated with the narrowed-down result is output in the form of print / display.
Therefore, even when a large amount of qualitative data that is unsuitable for quantification and has low searchability is analyzed, it can be extracted efficiently and appropriately. In addition, the data analysis that combines the quantitative data and the qualitative data can provide an effect that the marketing data meeting the needs of the customer can be provided accurately and easily.

Further, according to the present invention, the qualitative data is classified into one or a plurality of classification hierarchies, a predetermined classification code is added, the quantitative data search condition and the added classification code are input, and the quantitative data is searched. Quantitative data and qualitative data that meet the conditions and include the classification code are output as search results.
By configuring in this way, in addition to the search conditions related to the desired quantitative data, a large amount of analysis objects are narrowed down by the search conditions related to the code added to the qualitative data, and the quantitative data and qualitative data associated with the narrowed-down results are obtained. Output in the form of print / display.
Therefore, it is possible to set data search conditions that take into account the usage balance between the search conditions related to quantitative data and the search conditions related to qualitative data, and marketing data that meets customer needs can be provided accurately and easily. An effect is obtained.

Further, according to the present invention, the qualitative data is classified into a plurality of different hierarchies from the major classification to the minor classification, a predetermined classification code is added, and the quantitative data search condition and the minor classification to the minor classification are added. Any one or a plurality of classification codes up to are input, and quantitative data and qualitative data that satisfy the search conditions for the quantitative data and include the classification code are output as search results.
By configuring in this way, in addition to the search conditions related to the desired quantitative data, a large amount of analysis is performed with the search conditions related to the codes added to the qualitative data classified into different hierarchies in multiple ranges from large classification to small classification The target is narrowed down, and quantitative data and qualitative data associated with the narrowed down result are output in the form of print / display.
Therefore, it is possible to set data search conditions that take into account the usage balance between the search conditions for qualitative data classified into different levels in multiple ranges, from major classifications to minor classifications, and the search conditions for qualitative data. The marketing data can be provided accurately and easily.

Hereinafter, the best mode for carrying out the present invention will be described.
First, the basic concept of the present invention will be described. FIG. 1 shows a flow of service provision using a data analysis apparatus as an embodiment of the present invention. As shown in FIG. 1, the data analysis apparatus inputs data to be analyzed, performs data processing / analysis processing, and then outputs the data in an appropriate form as shown in FIG. Here, in the present invention, data to be analyzed is classified into so-called quantitative data and qualitative data, and respective databases are constructed. The quantitative data is data having data characteristics that can be quantified, and as a specific example, purchase history data (point card data, POS data) as shown in FIG. In the example shown in FIG. 2, the purchase history data includes a customer master including data such as a customer ID, name, age, sex, and address, a purchase history ID generated by a purchase transaction executed by the customer, a purchase date, and a purchase price. And attribute data composed of data such as
In addition, qualitative data is data with data characteristics that are difficult to quantify, and a specific example is a free answer questionnaire (eg; why did you choose our shop? This is the original response data.

The quantitative data is subjected to data processing such as decyl ranking and RFM ranking and is stored in a database (purchasing history DB). This situation will be described with reference to FIG. 2. The purchase history data (purchase history ID, customer ID, name, age, gender, address, purchase date, purchase price), for example, is aggregated by customer ID, and RFM rank. (R: Latest purchase date, F: Cumulative purchase count, M: Cumulative purchase price).
For example, a customer numbered with a customer ID: 00001 is assigned R rank 5 based on the latest purchase date, F rank 4 is assigned based on the cumulative purchase count, and M rank 4 is assigned based on the cumulative purchase price. Assigned.
These newly assigned RFM ranks are used as search conditions for customers such as the latest purchase date is new / old, the cumulative purchase count is large / small, and the cumulative purchase amount is large / small.

Similarly, as shown in FIG. 3, qualitative data is also subjected to data processing such as quantification by text mining and stored in a database (questionnaire DB). This situation will be described with reference to FIG. 3. Representative words are extracted from questionnaire originals of a plurality of answer questionnaires, and the plurality of questionnaire originals are classified into several groups depending on whether or not specific representative words are included. For example, one group is formed by original questionnaires including representative words (product, product lineup, number of products, abundant). A classification hierarchy that comprehensively indicates this representative word / phrase (hereinafter referred to as a small classification) is defined as a cluster (medium classification) for the large classification. For example, the cluster is selected from the representative classification (product, assortment, number of items). Furthermore, a code is added using a classification hierarchy that comprehensively indicates the cluster as a factor (major classification). For example, the factor is (product) from the cluster, and a code 001 is added.
In this way, factors, clusters, representative phrases, and questionnaire original text are categorized into a plurality of different levels in order from major classification to minor classification, and qualitative data (factors, clusters, representative phrases) other than the questionnaire original text are Coded in a phrase and used as a customer search criteria.

After such data processing, the qualitative data and quantitative data are integrated and stored in a database (integrated DB) by an identifier to be analyzed (customer ID). FIG. 4 shows a state in which the quantitative data stored in the purchase history DB and the qualitative data stored in the questionnaire DB are combined with the customer ID. As shown in FIG. 4, in addition to common items (customer ID, name, age, gender, address) of each data, RFM rank as quantitative data, Q1 major classification, Q1 medium classification, questionnaire original Q1 as qualitative data , Q2 (not shown) are combined to form a table (database).
The constructed integrated DB is searched from a terminal on the Internet by the ASP server. In other words, search conditions are set by gender, age, purchase store, time series, etc., and after grasping the overall trend from each aspect, that is, after narrowing down the customers to be searched In addition, it provides access to the original questionnaire (detailed data) associated with the refined search results.

  The above service provision is realized. Explaining this effect in comparison with the analysis of purchase history data (quantitative data, numerical data) only, the results obtained with the purchase history data analysis are limited to analysis of customer attributes and past purchase results. An analysis based on the fact of "what you purchased" is possible, but "what did you buy (satisfaction factor), what did you buy but what you are dissatisfied with (dissatisfaction factor), what do you expect? The customer ’s psychological aspect (causal relationship) was not reflected.

Complementing this is qualitative data. According to this service, data that reflects psychological aspects is obtained through questionnaires to customers, and this is integrated with purchase history data to clarify the reasons for purchasing customers. In addition, for each customer segment (segment separated by gender, age group, etc.) set by ranking, a material for planning a promotion measure considering psychological aspects can be provided.
Further, according to this service, the data analyst can quantitatively refer to the qualitative data classified by the text mining via the Internet, so that, for example, information sharing between departments is facilitated.

Hereinafter, an embodiment of a data analysis apparatus of the present invention will be described with reference to the drawings. FIG. 1 is a configuration diagram of a data analysis apparatus 1 of the present embodiment.
As described above, the data analysis apparatus 1 according to the present embodiment inputs quantitative data and qualitative data as analysis target data. Each of the input interfaces is the quantitative data input unit 10 and the qualitative data input unit 20.
That is, the quantitative data input unit 10 inputs quantitative data in a format associated with an identifier to be analyzed (hereinafter described as a customer ID). As a specific example, as described above, it is purchase history data such as POS data and point card data. As an input form, it is possible to input by reading data from an electronic medium in which these are recorded.
Further, the qualitative data input unit 20 inputs qualitative data associated with the customer ID. As a specific example, as described above, it is questionnaire data in a free answer format, and the input format may be a data transmission / reception mode or a manual input mode via a Web browser.
Data input to each input unit is stored in the quantitative data DB 11 and the qualitative data DB 21, respectively. That is, the quantitative data DB 11 stores quantitative data associated with the customer ID. Further, the qualitative data DB 21 stores qualitative data associated with the customer ID.

Quantitative data and qualitative data stored in these databases are processed using a rank assignment method that is appropriately selected to improve searchability and convenience. That is, the rank assigning unit 12 reads desired quantitative data from the quantitative data DB 11, analyzes the data using a rank assigning method selected after data cleansing, and sets a plurality of customers (customer IDs) as data items related to the quantitative data. Segmented as a search condition and stored.
Further, the text mining unit 22 reads desired qualitative data from the qualitative data DB 21 and classifies (codes) it by a selected text mining technique. Therefore, depending on the selected text mining technique, the text mining unit 22 classifies the qualitative data into one or a plurality of classification hierarchies and adds a predetermined classification code or a plurality of qualitative data from a large classification to a small classification. And processing for adding a predetermined classification code is executed.
These execution results are output to the data integration unit 30, respectively.

The data integration unit 30 inputs quantitative data and qualitative data from the rank assigning unit 11 and the text mining unit 21, respectively, and associates these quantitative data and qualitative data with a customer ID to create a data search table (database). To construct.
This data search table is read by the data search unit 40 and used for analysis of quantitative data and qualitative data. That is, the data search unit 40 inputs the search condition of the quantitative data from the quantitative data search condition part 41, and is linked to the quantitative data corresponding to the search condition and the quantitative data corresponding to the search condition by the data integration part. A search result including the qualitative data is output from the search result output unit 50. When the quantitative data search condition is input from the quantitative data search condition part 41 and the classification code added by the text mining part 22 is input from the qualitative data search condition input part 42, the data search part 40 The search result output unit 50 outputs the quantitative data and the qualitative data that correspond to the search condition and are linked by the data integration unit 30 including the classification code as a search result.

  As described above, the search conditions (including the classification code) input from the quantitative data search condition input unit 41 and the qualitative data search condition input unit 42 include the data characteristics, the data items created by the rank assigning unit 12, and the text mining unit 22. Although the data analyst can input arbitrarily according to the added classification code, in order to narrow down the qualitative data more efficiently, at least the search condition for quantitative data can be input. In this case, as described above, the qualitative data search condition input unit 42 does not require the input of a classification code as a search condition, and the search is performed so as to grasp the overall trend by sequentially increasing the search conditions for quantitative data. I do. Then, when a certain degree of narrowing is completed, the questionnaire original text data linked with the narrowed customer ID is output as it is based on the qualitative data search condition or without using the qualitative data search condition.

  When narrowing down by classification code, in addition to the quantitative data search condition input from the quantitative data search condition part 41, the data search part 40 further adds the large text mining part 22 added from the qualitative data search condition input part 42. Any one or a plurality of classification codes from classification to sub-classification are input, the quantitative data associated with the search condition of the quantitative data and associated with the data integration unit 30 including the classification code, and It may be configured to output qualitative data as a search result. In this case, one or a plurality of small classifications, medium classifications, and large classifications are used as a search condition depending on how much narrowing is desired to obtain a search result.

The quantitative data search condition input unit 41 and the qualitative data search condition input unit 42 are basically Web-based input interfaces, each of which can specify one or a plurality of search conditions.
Similarly, the search result output unit 50 is a Web-based output interface, and depending on the implementation, a search result transmission process to the data analyst's terminal, a display process on the data analyst's terminal, and a search result printing process Is executed.

Next, the operation of the data analysis apparatus 1 according to the present embodiment will be described with reference to the drawings. FIG. 6 is a flowchart showing a data analysis processing process of the data analysis apparatus 1 of the present embodiment when a purchase history is input as quantitative data and questionnaire data is input as qualitative data.
When the quantitative data input unit 10 and the qualitative data input unit 20 input quantitative data (number-based purchase history data) and quantitative data (text-based questionnaire data) regarding the customer to be analyzed (step S1 in FIG. 6), After each of the quantitative data DB 11 (purchase history DB) and the qualitative data DB 21 (questionnaire DB) is updated (step S2), as shown in FIGS. 2 and 3, the rank assigning unit 12 and the text mining unit 22 respectively. A new data item and classification code are added (step S3).

As shown in FIG. 4, each quantitative data and qualitative data output by the rank assigning unit 12 and the text mining unit 22 are input to the data integration unit 30, and are associated with the analysis target identifier (customer ID) as a key. The respective databases (tables) are integrated (step S4). As described above, a multidimensional analysis DB to which new data items and classification codes are added is constructed.
The identifier to be analyzed may be any identification information that can associate quantitative data and qualitative data, and may be associated with a plurality of identifiers. .

  The database integrated in this way is published on the Internet via a predetermined ASP server (step S5). When the ASP server is accessed from a data analyst's terminal on the Internet, and the quantitative data search condition and the qualitative data search condition are input to the quantitative data search condition input unit 41 and the qualitative data search condition input unit 42 (step S7, S9), the search unit 40 outputs the corresponding quantitative data and qualitative data (steps S8 and S10). As an output format (display format) at this time, as shown in FIGS. 7 to 11, it is conceivable to adopt a GUI-based format using an OLAP (OnLine Analytical Processing) tool.

Hereinafter, the process of searching the database integrated in this way will be described with reference to FIGS.
FIG. 7 shows that the quantitative data search condition input unit 41 specifies the age and sex analysis axes as the search conditions, and aggregates and outputs the customers stored in the integrated database shown in FIG. 4 using the age and sex analysis axes. The results are shown. In this case, the search condition (corresponding to the analysis axis of the expected requirement category shown in FIG. 7) is not specified in the qualitative data search condition input unit 42.
When the output result is further expanded for R rank and F rank, that is, when the analysis axis of R rank and F rank is designated as a search condition instead of gender, the search unit 40 integrates as shown in FIG. The customers stored in the database are aggregated and output by R rank and F rank analysis axes by age.

Now, looking at the output result shown in FIG. 8, when the data analyst specifies the analysis axis of the R-rank 5 and F-rank 5 expected factor categories (qualitative data) as search conditions, as shown in FIG. The search unit 40 is a factor in which R rank 5 and F rank 5 customers are further classified by the text mining unit 22 among the customers stored in the integrated database (major classification: price, product, service, salesperson). ) Aggregated and output on the analysis axis.
When the data analyst designates the analysis axis of the age and the sex (customer attribute information) once again for the output result, the search unit 40 aggregates with the analysis axis of the factor (major classification) classified by the text mining unit 22. The results are again aggregated and output on the age and gender analysis axes.
Looking at this, as an example, you can see that men in their 30s are expecting service.

  With respect to this output result, when the data analyst designates the analysis axis of the medium classification of R rank 5, F rank 5, 30s, male expected factor category as a search condition, as shown in FIG. Among the customers stored in the integrated database, R rank 5, F rank 5, 30s, male customers are further classified by the text mining unit 22 (medium classification: delivery, service corner, parking) (Fees, after-sales service, repair, warranty, sale, information provision) are aggregated and output.

With respect to the output result, when the data analyst specifies information provision among the factors (medium classification) classified by the text mining unit 22, the search unit 40 selects the quantitative data among the customers stored in the integrated database. R rank 5, F rank 5, 30s, male, qualitative data is medium classification: together with the customer corresponding to information provision, all the detailed data (qualitative data: questionnaire original text) are output as shown in FIG.
As described above, the customer's qualitative data corresponding to the desired rank and customer attribute and corresponding to the desired classification code is retrieved and output from the integrated database.

As described above, according to the data analysis apparatus 1 of the present embodiment, a large amount of analysis targets are narrowed down by a search condition relating to desired quantitative data, and qualitative data associated with the narrowed-down results are printed and displayed. Output.
Therefore, even when a large amount of qualitative data that is unsuitable for quantification and has low searchability is analyzed, it can be extracted efficiently and appropriately. In addition, the data analysis that combines the quantitative data and the qualitative data can provide an effect that the marketing data meeting the needs of the customer can be provided accurately and easily.

The data analysis apparatus 1 described above has a computer system inside.
A series of processes related to the data analysis process described above is stored in a computer-readable recording medium in the form of a program, and the above process is performed by the computer reading and executing this program.
That is, each processing means and processing unit in the data analysis apparatus 1 is performed by a central processing unit such as a CPU reading the above program into a main storage device such as a ROM or RAM and executing information processing / calculation processing. Is realized.
Here, the computer-readable recording medium means a magnetic disk, a magneto-optical disk, a CD-ROM, a DVD-ROM, a semiconductor memory, or the like. Alternatively, the computer program may be distributed to the computer via a communication line, and the computer that has received the distribution may execute the program.

Service provision flow diagram. FIG. 5 is a flow chart for determining the rank of quantitative data. Qualitative data coding example. The table block diagram which integrated quantitative data and qualitative data. 1 is a configuration diagram of a data analysis device 1. FIG. The flowchart which shows a data analysis process process. The figure which shows the search result in a customer attribute. The figure which shows the search result in a rank. The figure which shows the search result in a large classification. The figure which shows the search result in middle classification. The figure which shows the output result of detailed data.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 ... Data analyzer 10 ... Quantitative data input part 11 ... DB for quantitative data
12 ... Rank assignment unit 20 ... Qualitative data input unit 21 ... Qualitative data DB
22 ... Text mining unit 30 ... Data integration unit 40 ... Data search unit 41 ... Quantitative data search condition input unit 42 ... Qualitative data search condition input unit 50 ... Search result output unit

Claims (5)

First storage means for storing quantitative data associated with an identifier to be analyzed;
Second storage means for storing qualitative data associated with the identifier to be analyzed;
Data integration means for associating the quantitative data and the qualitative data with the identifier of the analysis target;
Data for inputting search conditions for the quantitative data and outputting search results including quantitative data corresponding to the search conditions and qualitative data associated with the quantitative data corresponding to the search conditions by the data integration unit A data analysis apparatus comprising: a search means. Text mining means for classifying the qualitative data into one or more classification hierarchies and adding a predetermined classification code;
The data search means inputs the quantitative data search condition and the classification code added by the text mining means, and corresponds to the quantitative data search condition and includes the classification code by the data integration means. The data analysis apparatus according to claim 1, wherein the linked quantitative data and the qualitative data are output as search results. The text mining means classifies the qualitative data from a large classification to a small classification into a plurality of ranges of different hierarchies, and adds a predetermined classification code,
The data search means inputs the search condition for the quantitative data and one or more classification codes from the large classification to the small classification added by the text mining means, and corresponds to the search condition for the quantitative data The data analysis apparatus according to claim 1, wherein the quantitative data and the qualitative data linked by the data integration unit including the classification code are output as search results. Data in a data analysis apparatus comprising first storage means for storing quantitative data associated with an identifier to be analyzed, and second storage means for storing qualitative data associated with the identifier to be analyzed An analysis method,
The quantitative data and the qualitative data are integrated by linking with the identifier of the analysis target,
Data for inputting the search condition for the quantitative data, and outputting a search result comprising quantitative data corresponding to the search condition and qualitative data linked to the quantitative data corresponding to the search condition Analysis method. Data is stored in a data analysis apparatus including first storage means for storing quantitative data associated with an identifier to be analyzed, and second storage means for storing qualitative data associated with the identifier to be analyzed. A program for executing analysis processing,
A process of linking and integrating the quantitative data and the qualitative data with the identifier of the analysis target;
To input a search condition for the quantitative data, and to execute a process of outputting a search result including quantitative data corresponding to the search condition and qualitative data linked to the quantitative data corresponding to the search condition Data analysis program.

Images