JP2005258774A - Questionnaire classification system and classification program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To classify questionnaires for supporting commodity planning and an advertisement activity meeting a life-style. <P>SOLUTION: A classification route preparation section 26 comprises: an analysis axis selection means for selecting an explanation variable falling in an analysis axis from the value of the explanation variable corresponding to questionnaire items on the basis of the questionnaire items included in answers to the questionnaires; a first classification rule storage means for calculating the coefficient of the explanation variable representing the weight of the questionnaire items for every analysis axis and storing the coefficient of the explanation variable as a first classification rule on a classification rule DB (Database) 18; a classification decision means for deciding the classification applied to a respondent for the questionnaires by cluster analysis on the basis of the values of the questionnaire items in accordance with the first classification rule; a center of gravity calculation means for calculating the center of gravity to the analysis axes of all the respondents applied to the classification; a second classification rule storage means for storing the center of gravity to each analysis axis for every classification as a second classification rule on the classification rule DB 18; and analysis result storage means for storing the classification applied to the respondent on a classification result DB 20. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

  The present invention relates to a system and program for classifying customer questionnaires.

  The company holds personal attribute information such as the customer's age, gender, and address as customer data, and holds the product name, price, purchase date, etc. as purchase history data for the product purchased by the customer. Customer data and purchase history data are analyzed and used for product planning and advertising activities.

When customer feedback and questionnaire results such as customer inquiries, requests, and complaints are analyzed and used for product planning and advertising activities, it is easy to analyze if customer feedback and questionnaire results are classified in advance. (For example, refer to Patent Document 1 and Patent Document 2). However, in order to perform the classification process, it is necessary to intervene by experienced persons (referred to as experts) with specialized knowledge.
JP 2003-76705 A JP 2003-58783 A Yutaka Tanaka, Kazumasa Wakimoto, "Multivariate Statistical Analysis" 8th Edition, Hyundai Mathematics Company, June 20, 1991, p.54-98, p.230-244 MRAnderberg, Hideo Nishida / Translation, “Cluster Analysis and Methods”, 2nd Edition, Uchida Otsukuru, December 20, 1995, p.209-227

  However, even if the customer data and purchase history data described above are analyzed, only the tendency based on age and gender can be grasped, and there is a case where the next action to be taken, for example, new product planning is not led. Or there exists a problem that the difference in the tendency by an age, sex, and an address cannot be found in an analysis result.

  An object of the present invention is to provide a questionnaire classification system and a classification program capable of classifying questionnaires without the knowledge of experts in order to support product planning and advertising activities tailored to lifestyles. .

  In the present invention, it is possible to provide information for a company to perform product planning and advertising activities in accordance with a customer's lifestyle. The outline of the present invention is as follows.

First, the following pre-processing is performed with an expert.
(1) The expert prepares lifestyle questionnaire items in the form of four choices (four choices) about the behavior of individual clothing, food and shelter. The person in charge conducts a questionnaire and collects the questionnaire results.

(2) An expert performs principal component analysis based on the questionnaire results, and calculates a small number of principal components (hereinafter also referred to as axes) and coefficients of explanatory variables. At this time, the system according to the present invention supports the work of an expert.

(3) Using the coefficients of the axes and explanatory variables, the expert performs a cluster analysis of the questionnaire results (ie, respondents) and groups the respondents. At this time, the system according to the present invention supports the work of an expert. Each group is called a classification i, and a generic name of the classification i is called a cluster. The cluster indicates lifestyle trends related to clothing (eg, fashionable, attractive), lifestyle trends related to food (eg, gourmet, health), and lifestyle trends related to living (eg, nature, mansion).

  The system according to the present invention stores the aforementioned axes and the coefficients of the explanatory variables as classification rule 1 (first classification rule), and stores the centroid of the aforementioned classification i as classification rule 2 (second classification rule).

  When the pre-processing is completed, when the person in charge conducts questionnaires on different subjects using the above-mentioned questionnaire items and stores the results, the system according to the present invention reads the questionnaire results and uses the classification rules 1 and 2 Then, a classification execution process for determining which of the classification i the questionnaire result belongs to is performed, and the classification result is output. This process can be repeated any number of times with different questionnaire targets.

  Thereafter, the person in charge can perform an analysis in consideration of the lifestyle in addition to the personal attribute and the purchase information based on the classification result and the customer information and purchase history data.

  In the patent document 1, the present applicant discloses a technique for performing classification based on the number of applications of words included in a customer's voice. However, according to this, there is a problem that the classification accuracy is lowered when there is little or no data matching the registered word. According to the present invention, since the same questionnaire item is used, data exists and it is possible to prevent a reduction in classification accuracy.

  More specifically, an object of the present invention is to select an explanatory variable serving as an analysis axis from the value of an explanatory variable corresponding to the question item based on the question item included in each of the questionnaire responses accumulated in the storage device. Analysis axis selection means and first classification rule storage means for calculating the coefficient of the explanatory variable representing the weight of the question item for each analysis axis, and storing the coefficient of the explanatory variable in the storage device as the first classification rule And, according to the first classification rule, based on the value of the question item included in the answer to the questionnaire stored in the storage device, class determination means for determining the class to which each of the respondents of the questionnaire belongs by cluster analysis Centroid calculating means for calculating centroids with respect to the analysis axes of all respondents belonging to the classification, and centroids with respect to each analysis axis for each of the classifications. Second classification rule storage means for storing in the storage device as a rule, and analysis result storage means for storing the classification to which the respondent determined by the classification determination means belongs in the storage device as an analysis result Achieved by a characteristic questionnaire classification system.

  In a preferred embodiment, the value of the respondent value for each category is further based on the average score of the respondent values for each question item and the average score of the respondent values belonging to each category for each question item. Answer value ratio generating means for generating a table showing the ratio, wherein the second classification rule storage means uses the identifier of the classification determined based on the ratio of the answer values as the second classification rule together with the classification It is configured to store in the storage device.

  More preferably, the analysis axis selection means includes eigenvalue calculation means for calculating each eigenvalue of each explanatory variable, and is configured to select an explanatory variable for which the calculated eigenvalue is greater than a predetermined value as an analysis axis. The

  In another preferred embodiment, a second class for determining a class to which each respondent of a new questionnaire for the question item belongs according to a first classification rule and a second classification rule stored in the storage device. Classification determination means.

  The second classification determining means, according to a first classification rule, for a value of a question item included in the answer to the questionnaire, a position calculating means for calculating a position of each answer on each analysis axis; Distance calculating means for calculating the distance between the position on each analysis axis and the center of gravity according to the classification rule of 2, respectively, wherein the classification with the smallest calculated distance is the classification of respondents of the questionnaire It may be determined.

  It is desirable that the value of the question item included in the questionnaire is a reply regarding clothes, food and / or living behavior of the respondent.

  Also, an object of the present invention is to select an analysis axis that selects an explanatory variable as an analysis axis from the value of an explanatory variable corresponding to the question item based on the question item included in each of the questionnaire responses accumulated in the storage device. Means for calculating the coefficient of the explanatory variable representing the weight of the question item for each analysis axis, and storing the coefficient of the explanatory variable in the storage device as the first classification rule; In accordance with the classification rule of the above, based on the value of the question item included in the questionnaire response accumulated in the storage device, the classification determination means for determining the classification to which each of the respondents of the questionnaire belongs, by cluster analysis, all belonging to the classification The center of gravity calculating means for calculating the center of gravity of the respondents with respect to the analysis axis, and the center of gravity for each analysis axis for each classification is described as a second classification rule. And causing the computer to function as analysis result storage means for storing the second classification rule storage means stored in the apparatus and the classification to which the respondents determined by the classification determination means belong as analysis results in the storage apparatus. It is also achieved by a computer-readable questionnaire classification program that is characteristic.

  In a preferred embodiment, the value of the respondent value for each category is further based on the average score of the respondent values for each question item and the average score of the respondent values belonging to each category for each question item. When the computer is made to function as an answer value ratio generating means for generating a table showing the ratio and the computer is made to function as the second classification rule storage means, it is determined based on the ratio of the answer values. The computer is operated so that the classification identifier is stored in the storage device together with the classification as a second classification rule.

  Further, in a preferred embodiment, when the computer is caused to function as the analysis axis selection means, the analysis axis selection means functions as eigenvalue calculation means for calculating each eigenvalue of each explanatory variable, and the calculated eigenvalue is a predetermined value. The computer is operated to select a larger explanatory variable as the analysis axis.

  In another preferred embodiment, the second classification determining means for determining the classification to which each respondent of the new questionnaire belongs according to the first classification rule and the second classification rule stored in the storage device. As described above, the computer is caused to function.

  More preferably, when the computer functions as the second classification determining means, each analysis axis of the answer is determined for the value of the question item included in the questionnaire answer according to the first classification rule. The computer is caused to function as a distance calculation unit that calculates a position between the position on each analysis axis and the center of gravity according to the second classification rule, and a position calculation unit that calculates the position above, In addition, the computer is operated so that the classification having the smallest calculated distance is determined as the classification of respondents to the questionnaire.

  If the person in charge conducts a questionnaire based on the above-mentioned questionnaire items and stores the result, the system according to the present invention uses the classification rule to perform the classification process, so that the classification result can be obtained without expert intervention. . The classification result can be used as effective information for performing product planning and advertising activities according to the lifestyle.

  That is, according to the present invention, to provide a questionnaire classification system and a classification program capable of classifying questionnaires without the knowledge of experts in order to support product planning and advertising activities in accordance with lifestyles. Is possible.

Embodiments of the present invention will be described below with reference to the accompanying drawings. FIG. 1 is a block diagram showing a configuration of a lifestyle questionnaire classification system (hereinafter simply referred to as “classification system”) according to an embodiment of the present invention. As shown in FIG. 1, the classification system 10 includes a support server 12 and a lifestyle questionnaire item DB 14 stored in an external storage device.
It has a lifestyle questionnaire result DB 16, a classification rule DB 18, a classification result DB 20, an input device 22 and a display device 24.

  The support server includes a classification rule creation unit 26 and a classification execution unit 28 as processing units for classification. The support system is not limited to being configured by a stand-alone computer. For example, a system may be constructed by connecting a support server, a terminal provided with an input device and a display device, and an external storage device provided with a DB via an electric communication line.

  When conducting a questionnaire for a customer (hereinafter, “customer” also includes a prospective customer), it is performed using a normal questionnaire question as shown in FIG. In the example of FIG. 2, the product to be purchased is an automobile, and the customer is inquired about points to be emphasized when purchasing an automobile.

  In this embodiment, in addition to the usual questionnaire questions as shown in FIG. 2, lifestyle questionnaire items (hereinafter also referred to as “questionnaire items”) of four-choice questions about the behavior of clothing, food and housing shown in FIG. ). Here, the four-choice question is for the respondent to select and answer one of the four choices (applicable, somehow apply, somehow do not apply, does not apply) to the question. .

  If the questionnaire questions include questions that ask the customer's personality and preferences (see, for example, Patent Document 2), the customer may feel uncomfortable with the surveying company because they feel that the inside of the individual is being looked into by a third party. And may show rejection in the answer.

  3-5 is a figure which shows the example of a questionnaire item. As shown in FIGS. 3 to 5, in the present embodiment, the question for grasping the lifestyle of the user from the three viewpoints of food (FIG. 3), clothes (FIG. 4), and residence (FIG. 5). Is included. In the present embodiment, a question that is received as an intention to grasp a psychological characteristic is not included in a questionnaire item so that the customer feels uncomfortable or does not cause a rejection reaction.

  For example, in FIG. 3, Question No. 101 “Even if the price is a little expensive, I often eat dishes using seasonal ingredients”, Question No. 102 “When I go out to eat, I choose a store with a coupon or a discount ticket. "There are many cases," and so on, to answer the customer's daily behavior. Hereinafter, the question No. is also simply referred to as “No.”.

  Next, the classification process performed by the support server will be described. FIG. 6 is a flowchart illustrating an example of processing executed in the classification rule creation unit 26 of the support server.

When a questionnaire using the above-mentioned questionnaire items is conducted for a predetermined number of customers (for example, 10,000 people) in advance, and the person in charge operates the input device 22 to input the collected questionnaire responses As shown in FIG. 7, the support server 12 stores a lifestyle questionnaire result (hereinafter, “questionnaire result”) in the lifestyle questionnaire result DB 16 of the external storage device in a format having a customer number and a numerical value of the answer to the question. Also stored as). In FIG. 7, the respondent of the customer number “NN168705” indicates that “2” is answered to the question No. 101 and “1” is answered to the No. 102.

  In the present embodiment, when the person in charge conducts and collects the questionnaire for the questionnaire items once set, and inputs and stores the results in the support server 12, the processing of the classification rule creation unit 26 described later is performed only once. Do. Thereafter, the questionnaire is executed and collected by using the same questionnaire items even for different subjects, and the classification execution unit 28 (to be described later) executes the process each time the result is input and stored in the support server 12. Will do.

  The processing of the classification rule creation unit 26 requires a large amount of calculation and requires a lot of processing time because it requires a lot of calculation and may require repetition by trial and error. On the other hand, the process of the classification execution unit 28 is an automatic process that does not require an expert and requires less processing time (usually less than about 1/10 of the process of the classification rule creation unit).

The processing of the classification rule creation unit 26 and the classification execution unit 28 of the support server will be described below using the questionnaire item “meal” (see FIG. 3) as an example.
[Process of classification rule creation part]
The processing performed by the classification rule creation unit will be described with reference to the flowchart of FIG.

[Principal component analysis (S10 in FIG. 6)]
Principal component analysis is based on the values of many explanatory variables (in this embodiment, the question number of the questionnaire item corresponds to this) (in this embodiment, the answer value of the questionnaire result corresponds to this), This is a method of compressing to a small number of indices (principal component = axis) with as little information loss as possible. A specific method of principal component analysis is described in Non-Patent Document 1, for example.

  The classification rule creation unit 26 reads the questionnaire result (see FIG. 7) from the lifestyle questionnaire result DB 16 and performs principal component analysis. FIG. 8 is a diagram illustrating an example of eigenvalues calculated by principal component analysis, and FIG. 9 is a diagram illustrating an example of coefficients of explanatory variables. The eigenvalue is the size of the information amount of the axis when the information amount of the explanatory variable is “1.0”. For example, in FIG. 8, the value “4.791” of the eigenvalue of “axis 1” represents that the information amount of the questionnaire item is 4.791 times as large as the information amount of each question. The axis with the larger eigenvalue is more important because the amount of information is larger, and the names “axis 1”, “axis 2”,.

  Generally, if there are p explanatory variables, the axis can be obtained up to “axis p”. However, the goal of principal component analysis is compression of explanatory variable amounts, and the original information can be explained sufficiently without using up to “axis p”. In general, an axis whose eigenvalue is “1” or more is adopted. . Also in this embodiment, an axis having an eigenvalue of “1” or more is adopted. Therefore, in the example of FIG. 8, “axis 1” to “axis 5” are employed.

  The coefficient of the explanatory variable represents the weight of the question. For example, in FIG. 9, question No. 101 has a positive weighting “0.30” for “axis 1”, and if the respondent answers a large value (for example, “4”) to No. 101, it is positive for “axis 1”. It shows that the relationship is strengthened. Also, since this question No. 101 has a negative relationship "-0.18" with "axis 2", it means that if the respondent answers a large value to this question, the relationship with "axis 2" weakens .

  The classification rule creation unit 26 stores the coefficient of the explanatory variable (FIG. 9) in the classification rule DB 18 of the external storage device as “classification rule 1 (first classification rule)”.

[Cluster analysis (S20 in FIG. 6)]
Based on “classification rule 1”, the classification rule creating unit 26 creates a classification by performing a cluster analysis on the questionnaire result (that is, respondent) by the K-means method.

  Cluster analysis is a method of automatically grouping respondents who answered in a similar pattern so that they belong to the same group. Each group is called “classification i (where i = 1, 2,..., N)”. The K-means method is often used in a plurality of cluster analysis methods (see, for example, Non-Patent Document 2). Note that it is necessary for a person to specify how many groups (that is, the number “n” in “classification n”), and the classification rule creation unit 26 receives the designation by an expert and performs processing.

  After determining the classification to which the respondent belongs, the classification rule creation unit 26 calculates the centroid of all the respondents belonging to “class i” (hereinafter referred to as “classification centroid”) and stores it in the external storage device. To do. FIG. 10 shows an example of the result obtained by performing the calculation in S20 for all the classifications, and shows the position of the center of gravity of each classification on each axis. For example, in the example of FIG. 10, the position of each axis is indicated as “−3.86” for the position of “axis 1” of the center of gravity of “class 1”, “0.25” for “axis 2”, and so on. .

[Calculation of response value ratio (S30 in FIG. 6)]
The classification rule creation unit 26 calculates the average score (AVT) of the answer values of the questionnaire results for each question number. Then, the average point (AVi) of the answer values of the questionnaire results belonging to the category i is calculated for each question No., and then the ratio of the answer values of the questionnaire results belonging to the category i for each question No. is calculated.

Response value ratio = AVi / AVT
For example, if the average score of “Category 1” of the question No.118 “I often go to fast food restaurants” is “2.50” and the average score of No.118 is “2.31”, The ratio of response values for “Category 1” is “2.50 / 2.31 = 1.08”. “Category 1” means that No. 118 is 8% higher than the overall average.

  The classification rule creation unit 26 calculates the ratio of answer values for all question numbers and classifications, and creates a table showing the ratio of answer values and a graph showing the ratio. FIG. 11 shows an example of a table of response value ratios, and FIGS. 12 to 14 show examples of graphs. The created table or graph is displayed on the screen of the display device 24 or printed from a printer according to the designation by the expert operating the input device 22. Experts analyze whether characteristics for each classification appear based on these tables and graphs.

  For example, referring to FIG. 12, the ratio of the answer value for the question “No.118“ I often go to fast food restaurants ”for“ Category 1 ”is high. "No.112" I often eat dishes that show the origin and brand of ingredients rather than the price ", No.113" Cooking using pesticide-free vegetables and additive-free ingredients " It is understood that there is a characteristic that the ratio of the answer value to the question “I often eat” is low. From such a result, the expert thinks that the respondent belonging to “Category 1” has a low consciousness regarding the item “food” and can be named “meal / non-careless” type.

  On the other hand, as for “Category 2”, No.118 “I often go to fast food restaurants” and No.102 “When I go out to eat, I often choose restaurants with coupons and discount tickets” It can be seen that there is a characteristic that the ratio of the answer value to No. 107 and No. 101 is low and the ratio of the answer value to No. 107 is low. Experts consider that respondents belonging to “Category 2” place importance on price for the item “food”, and can be named “deal”. The expert also examines and classifies other classifications, and if there is no similar classification and it is determined that each has unique characteristics, it is determined as OK and the process proceeds to step S40. Otherwise, steps S10 to S30 are executed again, and trial and error are repeated until it becomes OK (see S35).

  FIG. 15 shows an example of classification names that the experts have named when the items “clothes”, “food”, and “living” are classified as “6”. In addition, what is necessary is just to give a name arbitrarily so that an expert may understand the content of the classification easily.

[Reception of classification name and storage of classification center of gravity (S40 in FIG. 4)]
Upon receipt of the name assigned to each classification by the expert, the classification rule creation unit 26 associates the stored classification center of gravity (see FIG. 10) with “Classification rule 2 (second classification rule)”. Is stored in the classification rule DB 18 of the external storage device.

  An example of “classification rule 2” for the item “food” is shown in FIG. In FIG. 16, “Category 1” is given the name “Eating / Unattended”, indicating that the position on “Axis 1” is “-3.86”. Moreover, the example of the classification | category for every item to which the said customer belongs based on a customer's questionnaire result is shown in FIG. The classification result of each item for each customer is stored in the classification result DB 20 of the external storage device. For example, in FIG. 17, the classification of each item based on the questionnaire result of customer number “NN041023” is “Unubore” for the item “clothing”, “Health” for the item “Food”, and “Japanese style” for the item “Living”. ".

[Processing of classification execution department]
After the classification rule creation unit once stores the classification rule 1 and the classification rule 2 in the external storage device, the person in charge conducts and collects questionnaires on different subjects using the above-mentioned questionnaire items, and collects the data from the external When stored in the questionnaire result DB 20 of the storage device, the classification execution unit 28 of the support server executes the following classification process. FIG. 18 is a flowchart showing processing executed by the classification execution unit 28.

[Classification process (S110 in FIG. 18)]
The classification execution unit 28 reads “classification rule 1” and “classification rule 2” stored in the classification rule DB 18 of the external storage device and the customer questionnaire results in the lifestyle questionnaire result DB 16 and classifies the questionnaire results. .

  More specifically, first, the classification execution unit 28 calculates the position of a certain questionnaire result (that is, respondents) on each axis according to the following equation (1). (S120).

Respondent's axis j position =
Σ (coefficient of question k on axis j × answer value of question k) (1)
Here, the coefficient of the question k on the axis j is the value of the coefficient of the explanatory variable of “classification rule 1 (see FIG. 9)”, and the answer value of the question k is a value selected from four choices (four choices) Σ indicates that the number of questions is repeated.

  The classification execution unit 28 calculates the position of the respondent's axis j for all axes (in the present embodiment, from “axis 1” to “axis 5” in FIG. 9).

  A case where a respondent A makes an answer shown in FIG. 19 for the item “food” will be described as an example.

  For example, in FIG. 9, the coefficient of “No. 101” of “Axis 1” is “0.30”, and this is multiplied by the value “3” of the answer of No. 101 of respondent A (see FIG. 19). As described above, the calculation using the above-described equation (1) is performed for all the question numbers, and the sum of them is set as the position on the “Axis 1” of the respondent A. The same applies to “axis 2” to “axis 5”. FIG. 20 is a diagram illustrating calculation results of positions on the respective axes of “axis 1” to “axis 5” for the respondent A.

  Next, the classification execution unit 28 calculates the distance between the respondent's position and the classification center of gravity according to the equation (2) (S130).

The distance between the respondent's position and the center of gravity of “Class i” =
√ {Σ ((position of respondent's axis j − position of axis j of the center of gravity of classification i) ² )} (2)
Here, the position of “axis j” at the center of gravity of “classification i” is the value of “classification rule 2 (see FIG. 16)”, Σ indicates that the number of axes is repeated, and √ indicates the square root.

  The classification execution unit calculates the distance between the position of the respondent and the center of gravity of “classification i” for all classifications (in the present embodiment, “classification 1” to “classification 6”).

  For example, in the example of FIG. 10, the center of gravity of “Category 1” is “−3.86” for “Axis 1”, “0.25” for “Axis 2”, and so on.

Here, the position of the respondent A on “axis 1” is “10.06”, and the position on “axis 2” is “−0.01” (see FIG. 20).
(10.06-(-3.86)) ² + (-0.01-0.25) ² +…
Thus, the result obtained by calculating the square root of the sum of “axis 1” to “axis 5” is the distance from the classification (see FIG. 21). “Category i” having the smallest distance value is a category to which the respondent A belongs. According to the example of FIG. 32, the respondent A belongs to “Category 6 (Gourmet)”.

The classification execution department is responsible for all questionnaire results (ie respondents).
Equations (1) and (2) are calculated and stored in the classification result DB 20 of the external storage device as a classification result (see FIG. 16) (S140).

  As described above, the classification execution unit 28 of the support server uses the “classification rule 1” and “classification rule 2” to perform principal component analysis that requires a long calculation time and trial and error, and cluster analysis by the K-means method. It is possible to automatically determine the classification to which the respondent belongs and save the classification result in the classification result DB in a short time without performing the process and without using an expert.

  In addition, when the life cycle questionnaire item is changed, the classification rule creation unit 26 may execute the process again to create the classification rule. In order to cope with changes in customer behavior, it is desirable that the classification rule creation unit 26 creates a new classification rule every time a certain period elapses (for example, every year).

  After obtaining the classification results, the person in charge uses the classification results obtained in this embodiment, the answers to the usual questionnaire questions, the customer master created by another system, sales performance customer analysis, etc. Customer analysis can be performed by calculation programs and graph creation programs.

  The example of the customer analysis using the classification result acquired using the support server 12 concerning this Embodiment is shown.

  FIG. 22 shows the result of cross-tabulating the classification included in the item “food” and the ownership status of the digital camera. Referring to this result, it can be understood that the classification that does not own the most digital camera is “food / non-careless” and the classification that owns the most digital camera is “fashion”.

  FIG. 23 shows the result of cross-tabulating the classification included in the item “clothes” and the ownership status of the digital camera. From this result, it can be understood that respondents belonging to “Unubore” have a high ownership rate of digital cameras.

  FIG. 24 shows the result of cross-tabulating the purchase status of the digital camera by multiplying the classification of the item “food” and the “clothing” classification of the item “clothes”. That is, the ownership tendency of the classification of the item “food” when the classification of the item “clothing” is focused on a set of “Unubore” is shown.

  FIG. 25 is a graph showing the total number of age groups for “whether or not car maintenance is performed by a dealer” and the summation result of classification of lifestyle items “clothes”, “food”, and “living”. (Part). According to the aggregation by age group, only the tendency that customers with higher age groups are likely to receive maintenance appears (see FIG. 25 (a)), but lifestyle items “clothes”, “food” According to the classification of “residential”, customers who belong to “appearance” in the item “clothing”, “gourmet” in the item “food”, and “house” in the item “residential” have a ratio of receiving maintenance by the dealer. The characteristic that it is higher than the others clearly appears (see FIG. 25B).

The present invention is not limited to the above-described embodiment, and various modifications are possible within the scope of the invention described in the claims, and these are also included in the scope of the present invention. Needless to say, in the above-described embodiment, the support server 12 includes the classification rule creation unit 26 and the classification execution unit 28, and the program of the classification rule creation unit 26 is executed to store data (“classification” in the classification rule DB 18. "Rule 1" and "Classification rule 2") are stored. In addition, the program of the classification execution unit 28 is executed to apply the classification rules stored in the classification rule DB 18 and classify the respondents of the questionnaire from the questionnaire results including the answers of the questionnaire items. The present invention is not limited to such a form, and includes a case where the computer reads the program from a storage medium storing a program that causes the computer to function as a classification rule creation unit and a classification execution unit. Further, the configuration may be such that the classification rule creation unit 26 and the classification execution unit 28 are activated in accordance with an instruction from a terminal connected via a communication line. In this case, the questionnaire result DB may be connected to a terminal, and the data of the questionnaire result DB may be read by the server classification rule creation unit or the classification execution unit via a communication line.

FIG. 1 is a block diagram showing a configuration of a classification system according to an embodiment of the present invention. FIG. 2 is a diagram illustrating an example of a normal questionnaire item for a customer. FIG. 3 is a diagram showing an example of lifestyle questionnaire items used in the present embodiment. FIG. 4 is a diagram showing an example of lifestyle questionnaire items used in the present embodiment. FIG. 5 is a diagram showing an example of lifestyle questionnaire items used in the present embodiment. FIG. 6 is a flowchart showing a process executed by the classification rule creating unit according to the present embodiment. FIG. 7 is a diagram illustrating an example of a result of a lifestyle questionnaire. FIG. 8 is a diagram illustrating an example of eigenvalues calculated in the classification rule creation process. FIG. 9 is a diagram illustrating an example of the coefficient of the explanatory variable calculated in the classification rule creation process. FIG. 10 is a diagram illustrating an example of the centroid of the classification calculated in the classification rule creation process. FIG. 11 shows an example of a table of response value ratios obtained in the present embodiment. FIG. 12 is a diagram illustrating an example of a graph of answer values obtained in the present embodiment. FIG. 13 is a diagram illustrating an example of a graph of answer values obtained in the present embodiment. FIG. 14 is a diagram illustrating an example of a graph of answer values obtained in the present embodiment. FIG. 15 is a diagram showing an example of naming by experts for classification of each item in the present embodiment. FIG. 16 is a diagram illustrating an example of “classification rule 2” for the item “food” in the present embodiment. FIG. 17 is a diagram illustrating an example of classification for each item to which a customer belongs based on a questionnaire result. FIG. 18 is a flowchart illustrating processing executed by the classification execution unit according to the present embodiment. FIG. 19 is a diagram for explaining an example when a certain respondent A answers the item “food”. FIG. 20 is a diagram illustrating calculation results of positions on the respective axes of “axis 1” to “axis 5” for the respondent A. FIG. 21 is a diagram for explaining the distance between the position of the respondent and the center of gravity. FIG. 22 is a diagram illustrating a result of cross-tabulating the classification included in the item “food” and the ownership status of the digital camera. FIG. 23 is a diagram illustrating a result of cross-tabulating the classification included in the item “clothes” and the ownership status of the digital camera. FIG. 24 is a diagram illustrating a result of cross-tabulating the purchase status of the digital camera by multiplying the classification of the item “food” and the “clothing” classification of the item “clothes”. FIG. 25 is a graph showing the total number of age groups for “whether or not car maintenance is performed by a dealer” and the summation result of classification of lifestyle items “clothes”, “food”, and “living”. It is a figure which shows (part).

Explanation of symbols

10 Lifestyle Questionnaire Classification System 12 Support Server 14 Lifestyle Questionnaire Item DB
16 Lifestyle questionnaire result DB
18 Classification rule DB
20 Classification result DB
22 Input Device 24 Display Device 26 Classification Rule Creation Unit 28 Classification Implementation Unit

Claims (11)

Based on the question items included in each of the questionnaire responses stored in the storage device, an analysis axis selection means for selecting an explanatory variable serving as an analysis axis from the value of the explanatory variable corresponding to the question item;
A first classification rule storage means for calculating a coefficient of an explanatory variable representing the weight of a question item for each analysis axis, and storing the coefficient of the explanatory variable in a storage device as a first classification rule;
In accordance with the first classification rule, based on the value of the question item included in the answer to the questionnaire stored in the storage device, classification determination means for determining the class to which each of the respondents of the questionnaire belongs by cluster analysis;
Centroid calculating means for calculating centroids of all the respondents belonging to the classification with respect to the analysis axis;
Second classification rule storage means for storing the center of gravity of each analysis axis for each analysis axis in the storage device as a second classification rule;
A questionnaire classification system, comprising: analysis result storage means for storing the classification to which the respondent belongs determined by the classification determination means in the storage device as an analysis result. Furthermore, based on the average score of respondents for each question item and the average score of respondents belonging to each category for each question item, a table showing the proportion of respondent values for each category is generated. Answer value ratio generating means to
The second classification rule storage means is configured to store, together with the classification, a classification identifier determined based on the ratio of the answer values in the storage device as a second classification rule. The questionnaire classification system according to claim 1. The analysis axis selection means includes eigenvalue calculation means for calculating each eigenvalue of each explanatory variable, and is configured to select an explanatory variable for which the calculated eigenvalue is greater than a predetermined value as an analysis axis. The questionnaire classification system according to claim 1 or 2. And a second classification determining means for determining a classification to which each respondent of the new questionnaire for the question item belongs according to the first classification rule and the second classification rule stored in the storage device. The questionnaire classification system according to any one of claims 1 to 3, wherein: The second classification determining means;
In accordance with the first classification rule, with respect to the value of the question item included in the answer to the questionnaire, position calculating means for calculating the position of each answer on each analysis axis;
Distance calculating means for calculating the distance between the position on each analysis axis and the center of gravity according to the second classification rule,
The questionnaire classification system according to claim 4, wherein the classification having the smallest calculated distance is determined as a classification of respondents to the questionnaire. The questionnaire classification according to any one of claims 1 to 5, wherein a value of a question item included in the questionnaire is a reply about clothes, food, and / or living behavior of respondents. system. Based on the question items included in each of the questionnaire responses stored in the storage device, an analysis axis selection means for selecting an explanatory variable serving as an analysis axis from the value of the explanatory variable corresponding to the question item,
A first classification rule storage means for calculating a coefficient of an explanatory variable representing a weight of a question item for each analysis axis, and storing the coefficient of the explanatory variable in a storage device as a first classification rule;
In accordance with the first classification rule, classification determination means for determining a classification to which each of the respondents of the questionnaire belongs by cluster analysis based on the value of the question item included in the questionnaire response accumulated in the storage device,
Centroid calculating means for calculating the centroid of all respondents belonging to the classification with respect to the analysis axis;
Second classification rule storage means for storing the center of gravity of each analysis axis for each analysis axis in a storage device as a second classification rule; and
A computer-readable questionnaire classification program that causes a computer to function as analysis result storage means for storing the classification to which the respondent belongs determined by the classification determination means in the storage device as analysis results. Furthermore, based on the average score of respondents for each question item and the average score of respondents belonging to each category for each question item, a table showing the proportion of respondent values for each category is generated. And functioning the computer as an answer value ratio generating means, and
When the computer functions as the second classification rule storage unit, the classification identifier determined based on the ratio of the answer values is stored in the storage device together with the classification as the second classification rule. 8. The questionnaire classification program according to claim 7, further comprising operating the computer. When the computer is made to function as the analysis axis selection unit, the analysis axis selection unit functions as an eigenvalue calculation unit that calculates each eigenvalue of the explanatory variable, and
9. The questionnaire classification program according to claim 7, wherein the computer is operated so as to select an explanatory variable for which the calculated eigenvalue is larger than a predetermined value as an analysis axis. Further, the computer is caused to function as a second classification determining means for determining a classification to which each respondent of a new questionnaire belongs according to the first classification rule and the second classification rule stored in the storage device. 10. The questionnaire classification program according to any one of claims 7 to 9, wherein: When operating the computer as the second classification determining means,
In accordance with the first classification rule, for the value of the question item included in the answer to the questionnaire, a position calculating means for calculating the position of each answer on each analysis axis, and
In accordance with the second classification rule, the computer is caused to function as distance calculation means for calculating the distance between the position on each analysis axis and the center of gravity, and
The questionnaire classification program according to claim 10, wherein the computer is operated so as to determine the classification having the smallest calculated distance as the classification of respondents to the questionnaire.

Images