JPH10171777A - System for analyzing data - Google Patents

Abstract

PROBLEM TO BE SOLVED: To automatically discover a characteristic category from among plural categories, and to present it to a user. SOLUTION: A category classifying means 0102 defines the group of a record constituted of more than one item stored in a disk 0101 as objective data, defines more than one items as a category item, and classifies the group of the record into plural categories according to the value of the category item. A category evaluating means 0106 calculates the total amounts of each of plural categories 0103 by a preliminarily decided calculation method, and calculates the evaluated value of each of the plural categories based on the total amounts. A category selecting means 0104 selects one or plural categories from among the plural categories based on the evaluated value, and a presenting means 0105 presents the selected category to a user.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a data analysis technique for classifying a target data into various categories and examining characteristics of each category for a large amount of data.

[0002]

2. Description of the Related Art As an analysis-oriented data utilization technique, OL
APs and multidimensional databases are known. These are described in "DATABASE SYSTEM" Vol.2 / No.3 (SoftBank Corp., 1996) and special feature "Practice of Data Warehouse". As a technique for analyzing a large amount of data and discovering knowledge, there is a technique for automatically extracting characteristics of data in Japanese Patent Application Laid-Open No. 8-77010 "Data analysis method and apparatus". Also,
These techniques are described in "Knowledge Discovery Techniques from Databases" (System / Information / Control, Vol. 39 / No. 4).

[0003]

[0004] Techniques such as a multi-dimensional database have made it possible to analyze target data from multiple angles. According to these analysis techniques, target data can be easily classified into various categories, and characteristics of each category can be examined. However, it is necessary for the user to specify which category to focus on, and when there are many possible categories, it is difficult to find a characteristic category from the categories.

An object of the present invention is to automatically find a characteristic category from a large number of categories and present it to a user.

[0005]

In order to solve the above problems, the present invention provides a set of records including one or more items as target data, one or more of the items as category items, A category classifying means for classifying the set of records into a plurality of categories according to the value of the item; calculating a statistic for each of the plurality of categories by a predetermined algorithm; and calculating the statistics of the plurality of categories based on the statistic. Category evaluation means for calculating an evaluation value for each, category selection means for selecting one or a plurality of categories from the plurality of categories based on the evaluation value, and presenting the selected category to a user Presenting means is provided.

Further, there is provided means for displaying statistics of one or more selected categories in a table format.

The apparatus further comprises means for setting one or more categories as a reference by a user, wherein the category classification means includes the set of records in the category as the reference according to the value of the category item. Classification into multiple categories.

The apparatus further comprises means for setting one or more categories selected by the category selecting means as a new reference category, setting a reference category, classifying the category, calculating an evaluation value, Each process of selecting a category and setting a new reference category is repeatedly executed.

In addition, there is provided means for storing a set of category items and fragments of explanatory texts regarding statistics, and the selected item is selected based on the value of the category item for one or more categories selected by the category selecting means. Means for calculating a statistic of one or more categories from the target data, and combining the statistic and the fragment of the description according to the value of the category item; Means for synthesizing an explanatory text explaining the feature of the quantity and means for displaying the explanatory text are provided.

[0010] Further, holding means for holding category item values and category evaluation values for one or more selected categories is provided, and the means for displaying the statistics in a tabular form comprises: Is displayed, means for highlighting an element of a table corresponding to a category having a maximum or minimum evaluation value among one or a plurality of selected categories is provided.

A category selection result holding means for holding the one or more categories sequentially selected by the category selection means in the repetition execution processing while maintaining a selected order; A means is provided for displaying the contents held by the category selection result holding means in a list format according to the order selected during execution or after the end of the repetitive execution processing.

In the repetitive execution process, a statistic of one or more categories selected at a certain time is presented to a user during the execution of the repetitive execution process, and at the same time, Means are provided for presenting the user with the selected one or more categories.

A category selection result holding unit for maintaining the one or more categories sequentially selected by the category selection unit in the repetition execution process while maintaining a selected order; After the end, means for presenting the contents held by the category selection result holding means to the user in the selected order is provided.

Also, the first data communication path connected by the data transfer path
And a second device, the first device comprising: means for generating a plurality of categories based on item information relating to target data; and a plurality of categories based on the generated category information and a statistic of each of the plurality of categories. Evaluation value calculating means for calculating an evaluation value for each of the categories, category selecting means for selecting one or more categories from the plurality of categories based on the evaluation value, and providing the user with the selected category. The second device includes a storage unit configured to store a set of records including one or more items as target data, and a statistic for each of the plurality of categories by a predetermined algorithm. A statistic calculating means for calculating, and a statistic holding means for holding a statistic calculated by the means, wherein the statistic calculating means of the second device comprises: The statistical information is calculated based on the category information transferred from the first device via the data transfer path and the target data in the storage unit, and the calculated statistical value is transmitted to the first device via the data transfer path. The evaluation value calculation means of the first device calculates the evaluation value based on a statistic transferred from the second device via a data transfer path.

Further, the processing procedure in the data analyzer is stored in a storage medium as a computer program.

[0016]

FIG. 1 shows an embodiment of the present invention. The target data 0101 is a set of records, and each record includes one or more items. FIG. 2 shows an example of the target data. This is an example of simple customer data. One or more of the prefecture, age, gender, occupation, and usage amount of these items are used as category items (for example, prefecture, age, gender, occupation) in the category classification means. FIG. 3 shows the flow of processing in the category classification means. First, among the items of the target data, the item used as the category item is designated (0301). A set of category items is {Ii} (i = 1 to M). The number of category items is M. Category item
A set of category values (for example, Tokyo, Kanagawa, self-employed, etc.) that Ii can take is defined as {Vij} (j = 1 to Ni). The number of possible category values Ni differs for each category item. For example, the number of possible category values for the category item “sex” is two for “male” and “female”. Next, a special category value “don't-care” is added to the set of possible category values for all category items (0302). As a result, the set of possible category values of category item Ii is {Vij} (j = 0 to Ni, where Vi0 is "don't-car
e ”). Next, a category is generated (0303). A category is defined by a combination of category item values, including “don't-care”. FIG. 4 shows an example of the category. For example, a category in which the category value of all category items is “don't-care” represents the entire set. One of the category items, the category value of "occupation" is "self-employed"
The category whose other category item is “don't-care” represents a set of self-employed persons. Next, all records of the target data are classified into the category generated in the process 0303 (0304). The category item value of each record is compared with the category item value of each category. If all the category item values match, the record is classified into that category. However,
"Don't-care" shall match all item values. Therefore, as shown in FIG. 5, one record is classified into a plurality of categories. In FIG. 5, the value of the category item “prefecture” is “Tokyo”, the value of the category item “age” is “20”, and the category item “gender”
Is "male" and the category item "occupation" is "employee"
Are classified into a plurality of categories such as a “male category” and a “company employee category”. However,
For the category of male self-employment, the category item Occupation
Are not classified because their values do not match. This is performed for all records (0305). The above classification process is performed for all possible categories (0306). In this way, a category and a set of records included therein are determined.

The case where the category items are independent has been described above. Next, the case where there is a dependency between the category items will be described. FIG. 6 shows an example of data having a dependency between category items. The possible category values for the category item "Region" are "Tohoku" and "Kanto"
"Chubu" etc. The possible category value of the category item “prefecture” is the name of the prefecture. These category items are not independent of each other and have a hierarchical structure as shown in FIG. Therefore, at the time of category generation, meaningless categories such as "the region is Tohoku and the prefecture is Okinawa" are excluded using the information of this hierarchical structure. For this purpose, as shown in FIG. 8, the category classification means inputs hierarchical structure information (0802) in addition to the target data.

In some cases, a category is generated under certain conditions regarding category items. An example of a condition regarding a category item will be described. As mentioned earlier, categories are defined by a combination of category item values, including "don't-care". Categories with more "don't-care" are more comprehensive categories, and more category values that are not "don't-care" are more detailed categories.
An inclusion relationship as shown in FIG. 9 is formed. For example, the “male employee category” is included in the “employee category”. At the same time, the “male office worker category” includes the “20s male office worker category”. Now, it is assumed that the user's interest is in each of the categories of “male office worker” and “female office worker” shown in FIG. 10, and a more detailed analysis of these categories is to be performed. In such a case, these categories are set as “reference categories”. In the case of FIG.
Since a plurality of categories of “male office worker” and “female office worker” are set as reference categories, they will be referred to as “reference category group”. The category classification unit 0102 (FIG. 1) generates only a more detailed category included in the reference category. In other words, "don '
The value of the category item that is not “t-care” is not “don't-care” even in the generated detailed category, and inherits the category item value in the reference category. In the example of FIG. 10, in the reference category, the category value of the category item “sex” is “male” and “female”, and the category value of the category item “occupation” is “company employee”. In the detailed category group, these category values are inherited, for example,
The category value of the category item “sex” does not become “don't-care”, and the category value of the category item “occupation” does not become “self-employment”. In the reference category,
The category item "age" which was "don't-care" has a category value other than "don't-care" such as "20s" in the detailed category.

In the category evaluation means 0106 (FIG. 1),
An evaluation value is calculated for each of the categories generated by the category classification unit 0102. The evaluation value is calculated from statistics for each category. An example of a statistic for a category is a statistic for a value of a specific item of the target data other than the item selected as the category item. This particular item, which is the subject of the statistics, will be referred to as "target item". For example, in the data example shown in FIG. 2, the item “usage amount” can be used as a target item. As the statistic, the sum and average of the “usage amount” are used. In addition, an item that does not take a numerical value such as the item “sex” can be set as a target item. In this case, as a statistic, a value of the target item in each category is a specific value, for example, a frequency of “male” is used. Further, the number of records included in the category may be used as the statistics of the category.

Based on the difference between the expected value of the statistics and the actual value, an evaluation scale of each category is defined by the following equation. Here, the realization value is a value obtained based on the value of the target item, and for example, 1100 in the case of a man / twenty and 700 in a case of a woman / 20s in FIG. The expected value will be described later. Category evaluation value μ
Is represented by the following equation. μ = {(A−E) · (A−E)} / (EE) where μ is the evaluation value of the category, A is the actual value of the statistic about the category, and E is the statistic of the statistic about the category. Expected value. Using the category group shown in FIG. 11 as an example,
The evaluation scale will be described. First, a reference category group is set, and a detailed category included in the reference category is generated. In FIG. 11, “male office worker” and “female office worker” are set as a reference category group, and “30-year-old male office worker” and “30-year-old female office worker” are generated as detailed category groups. In the table described as “realized value”, the realized value of the statistic in the detailed category group is shown, and in the table described as “expected value”, the expected value of the statistic in the detailed category group is shown. Here, the total value of the values of the target items is used as the statistic. The value of the statistic of the category Ci in the reference category group is Si (where i is a subscript corresponding to each category), and the total value of the statistic in the reference category group is S0.
And In FIG. 11, if the category of “male office worker” is C1 and the category of “female office worker” is C2, S1 = 6
800, S2 = 4100, S0 = 10900. Also, the realization value of the statistics of the detailed category Di is SDi, the expected value is EDi,
The sum of the actual values of the statistics in the detailed category group
SD0. However, the category Di is the category C
The detail category included in i. At this time, the expected value is given by the following equation. EDi = Si · (SD0 / S0) In FIG.
1. If the category of "30-year-old female office worker" is D2, S
D1 = 1800, SD2 = 1600, SD0 = 3400, ED1 = 2121
(S1 / (SD0 / S0) = 6800 / (3400/10900)), E
D2 = 1279 (S2 · (SD0 / S0) = 4100 · (3400/1090
0)). Further, when the evaluation value μ is calculated by the above equation using the actualized value and the expected value of the detailed category obtained as described above, μ = μ (1800−2121) · (1800 −2121)｝ / (2121 ・ 2121) =
0.023 For “30-year-old female employee,” μ = {(1600-1279) ・ (1600-1279)} / (1279 ・ 1279) ＝
0.063.

In the category selecting means 0104 (FIG. 1), there are a case where a category is selected alone and a case where a category is selected as a category group. When selecting a category alone, one category having the largest evaluation value or a plurality of categories is selected from the generated categories in descending order of the evaluation value. When selecting as a category group, the evaluation value of the category group is calculated, and based on this, one category group having the largest evaluation value or a plurality of category groups are selected in descending order of the evaluation value. A set of categories included in a certain category group G is represented by {C i} (i = 1 to n),
Assuming that the evaluation value of the category Ci is μi and the evaluation value of the category group G is μG, μG is defined by the following equation. μG = max (μi) (where i = 1 to n) In this case, the maximum evaluation value of each category included in the category group is set as the evaluation value of the category group. Also, the definition of μG can be given by the following equation. μG = Σμi (where i = 1 to n) In this case, the sum of the evaluation values of the categories included in the category group is used as the evaluation value of the category group. Also, μG
Can be given by the following equation.

ΜG = Σμi / n (where i = 1 to n) This means that the average of the evaluation values of the categories included in the category group is used as the evaluation value of the category group.

The presenting means 0105 (FIG. 1) presents the category or categories selected by the category selecting means 0104 to the user. FIG. 12 shows an example in which a list of a plurality of selected categories is presented to the user. The number of categories to be displayed in the list is specified by the user. At this time, if the number of categories is designated as 1, one selected category is presented. From the category selecting means 1201 (corresponding to the category selecting means 0104 in FIG. 1), the designated number of categories are selected in descending order of the evaluation value, and are presented as a category list 1202 to the presenting means 1203 (corresponding to the presenting means 0105 in FIG. 1). ). The category list 1202 stores values of category items in each category. Presentation means 120
In step 3, the contents of the category list 1202 sent from the category selection means 1201 are presented to the user in the form of a list (1204).

FIG. 13 shows an example in which a list of a plurality of selected category groups is presented to the user. The number of categories displayed in the list is specified by the user. At this time, if the number of category groups is designated as 1, one selected category group will be presented. From the category selecting means 1301 (corresponding to the category selecting means 0104 in FIG. 1), a designated number of category groups are selected in descending order of evaluation value, and presented as a category group list 1302 to the presenting means 1303 (presenting means 0105 in FIG. 1). Corresponding to). The category group list 1302 stores the values of category items in each category group. Presentation means 1
At 303, the contents of the category list 1302 sent from the category selection means 1301 are presented to the user in the form of a list (1304). However, the display of the category item that inherits the value in the reference category group can be omitted. In the case of FIG. 13, assuming that “male office worker” and “female office worker” are the reference category group, only “age = 30” is displayed for the first category group in the category group list. The category group may be presented in a tabular format. FIG.
Is a table of category item values in a certain category group and the corresponding statistics.

FIG. 15 shows the structure of a means for presenting a table. Category selecting means 1502 (corresponding to category selecting means 0104 in FIG. 1) to presenting means 1504
The list 1503 of the category group sent to (corresponding to the presentation means 0105 in FIG. 1) is only information on the category item values, and the values in the table, that is, the values of the statistics in each category refer to the category item values. While, it is calculated from the target data 1501 (corresponding to the target data 0101 in FIG. 1). That is, the records classified into the selected category group are searched from the target data 1501 while referring to the category item values (1505), and the statistic is calculated for these records (1506). In addition, the categories that the user should pay attention to, such as the category having the highest evaluation value among the categories included in each category group, are highlighted (see “21 in the“ 30s female category ”in FIG. 14).
0 ").

For this purpose, as shown in FIG. 16, in addition to the information on the category item values in the category group, the information on the category item values relating to the category of interest is added to the category selection means 1602 (corresponding to the category selection means 0104 in FIG. 1). )
To presentation means 1604 (corresponding to presentation means 0105 in FIG. 1). In this example, a list 1603 of category groups
In the above, information indicating that the feature value of the “female in thirties” category is “+60” is sent as the focused category information. The feature amount in this case means that the deviation of the actual value of the statistic from the expected value is + 60%. These tabular displays are activated when the user designates an arbitrary category group from the category group list when a list of category groups as shown in FIG. 13 is presented to the user. You.

There is also a method of presenting a user with a sentence explaining the characteristics of the category and the category group. This example is shown in FIG. For this purpose, similarly to the example of FIG. 16, the feature amount of the category is added together with the category item information from the category selection unit 1701 (corresponding to the category selection unit 0104 in FIG. 1) to the presentation unit 1703 (corresponding to the presentation unit 0105 in FIG. 1). ). The presenting unit 1703 combines the explanatory template 1704 prepared in advance in accordance with the feature amount and presents it to the user in the form of an explanatory note 1705. In this example, templates such as “somewhat high”, “very high”, and “somewhat low” are prepared, and “somewhat high” is set according to the feature amount “+60” of the target category in the category group list 1702. It is selected and assembled in the form of an explanatory note with category item values. The individual processing means and their alternative means have been described above.

Next, the flow of processing combining these means will be described. FIG. 18 shows this example. Information to be input includes target data 1801 and dimension information 1802
It is. The dimension information 1802 includes item information indicating which of the items of the target data 1801 is a category item and which is a target item, and information on a hierarchical structure between the category items. Also, a reference category group is set (1804). The first reference category group setting is performed by the user 1803. Category classification means 1805
Then, a detailed category included in the reference category is generated. At this time, in each detail category, one of the category values “don't-care” in the reference category
Only one category item has a value other than "don't-care".
This is shown in FIG. When “male self-employment” is used as the reference category, a value other than “don't-care” is set as one of “prefecture” and “age” as the detailed category. "Don't-ca" for both "prefecture" and "age"
Do not set a value other than "re". Records of the target data are classified into these detailed categories 1806.

The category evaluation means 1815 calculates a statistic 1807 and an evaluation value 1808 for each detailed category 1806. The category selecting unit 1809 evaluates and selects these detailed categories as a category group, and the presenting unit 1810 presents a list of the category groups to the user. Users see this list,
It is possible to know a category group to be focused on. Further, when the user designates an arbitrary category group from this list (1811), a table of statistics in this category group is displayed (1814). Similarly, an explanatory sentence describing the statistic feature in this category group is displayed (1813). Further, the user designates an arbitrary category group from the list as the next reference category group, so that more detailed analysis of the categories included in the reference category group can be continuously performed.

In FIG. 18, after the category is selected, instead of the user designating the next reference category group, the top category group in the category group list is automatically set as the next reference category group. Analysis of various categories can be performed automatically and continuously. FIG. 20 is a modification of the processing after the category selecting means in FIG. 18 so that detailed category analysis can be automatically continued. Information on the category group obtained as a result of the category selection 2001 is sent to the selection result holding unit 2002.
It is also sent to the reference category setting means at the same time. When the next reference category group is set by the reference category setting means, the analysis of the detailed category is automatically started. The selection result holding unit 2002 sequentially receives and holds the results of the category selection performed automatically and continuously. Also,
The held category selection results can be read out in the order in which the category selections were made, or in any order, at any time. Thus, while the new detailed category analysis is being performed, the user can read out the information of the group of categories obtained as a result of the previous analysis from the selection result holding unit 2002, and read the information in the tabular form 2004 or the process On display 2003, it can be seen.

FIG. 21 shows an example of display in the table format display means 2004. The tabular display unit 2004 displays the statistics in the category group in an arbitrary stage of the automatic continuous analysis in a tabular format. If the analysis result of the next stage has already been obtained, the category item value added in the next stage is displayed at the same time. In FIG. 21, "Next category"
“Occupation = company employee” displayed as “” is a category item value to be focused on in the next stage. If the next stage of analysis is currently in progress and results are not yet available, notify the user that analysis is in progress and, when results are obtained, add the added category item values. indicate. An explanatory note describing the characteristics of the category group can be displayed together with a table.

FIG. 22 shows an example of a display on the process display means 2003 in FIG. In the process display means 2003, the category items of the category group selected up to the present, starting from the first reference category group, and their item values are sequentially displayed. In FIG. 22, "sex" and "age" are category items for the first reference category group. In the first analysis stage, the value is "don't-
Of the category items that were "care", "prefecture = Tokyo" was selected, and the second analysis stage indicates that this was used as a reference category group and "occupation = company employee" was selected. If the user explicitly selects "occupation = self-employed" in the first analysis stage, in the second analysis stage,
The analysis is continued with this as a reference category group. If an analysis is in progress, these displays are updated each time an analysis result is obtained. When the user designates an arbitrary category group on the process display, information on the designated category group can be displayed on the tabular display means. In addition, the tabular display means 2004 can sequentially reproduce the steps of the automatic continuous analysis. That is, a table for the category group displayed on the process display means 2003 is displayed in the order of the processes. By the operation of the user, it is possible to proceed to the category group of the next stage or to return to the previous stage.

Next, FIG. 23 shows an example in which analysis processing is performed in a distributed computer environment. A computer 2301 that stores target data is provided separately from a computer 2302 that implements category generation, category selection, presentation, and the like. Category generation means 2
In step 307, the information of the category item about the generated category is transmitted to the computer 2301 that holds the target data. The computer 2301 holding the target data searches the target data 2303 based on the received information on the category item, and calculates the statistics of the category (230).
4). Then, only the calculated statistics are sent back. The computer 2302 that performs the analysis calculates an evaluation value using this statistic (2308), and selects a category (230).
9). According to the above method, the computer 2302 that performs the analysis
In addition, the communication amount can be reduced as compared with the case where the target data itself is transmitted. If the calculated statistics are stored in the computer 2301 that stores the target data, (23
05), the time to calculate the statistic can be reduced the next time the same analysis is performed. Further, when the target data is shared with a plurality of analysis computers, the required statistics are already calculated according to a request from another analysis computer, and the time for calculating the statistics may be reduced in some cases.

[0034]

According to the present invention, in a data analysis of a large amount of data, a combination of a large number of items that cannot be handled manually is automatically searched to find a characteristic category to be noticed, It can be presented to the user.

[Brief description of the drawings]

FIG. 1 is a diagram showing a configuration of one embodiment of the present invention.

FIG. 2 is a diagram illustrating an example of target data.

FIG. 3 is a diagram showing a flow of a category classification process.

FIG. 4 is a diagram illustrating an example of a generated category.

FIG. 5 is a diagram showing how records are classified into a plurality of categories.

FIG. 6 is a diagram showing an example of data having a hierarchical structure between items.

FIG. 7 is a diagram illustrating an example of a hierarchical structure of category values.

FIG. 8 is a diagram showing a configuration diagram of an embodiment using hierarchical information.

FIG. 9 is a diagram showing an inclusion relationship between categories.

FIG. 10 is a diagram showing a relationship between a reference category and a detail category.

FIG. 11 is a diagram showing an example of a realized value, an expected value, and an evaluation value of a statistic.

FIG. 12 is a diagram showing a configuration of means for presenting a list of categories.

FIG. 13 is a diagram showing a configuration of means for presenting a list of category groups.

FIG. 14 is a diagram showing an example of a tabular display of statistics.

FIG. 15 is a diagram showing a configuration of a unit for performing a tabular display of statistics.

FIG. 16 is a diagram showing a configuration of a means for highlighting a focused category.

FIG. 17 is a diagram showing a configuration of a means for displaying an explanatory note.

FIG. 18 is a diagram showing a configuration of another embodiment of the present invention.

FIG. 19 is a diagram illustrating an example of a detailed category generated from a reference category.

FIG. 20 is a diagram showing a configuration of a means for performing automatic continuous processing.

FIG. 21 is a diagram showing an example of a tabular display of statistics in automatic continuous processing.

FIG. 22 is a diagram illustrating an example of a process display in the automatic continuous process.

FIG. 23 is a diagram showing a configuration using distributed computers.

[Explanation of symbols]

 0101, 0801 Target data 0102, 0803 Category classification means 0103, 0804 Category 0104, 0805 Category selection means 0105, 0806 Presentation means 0106, 0807 Category evaluation means 0802 Hierarchical information

Continuing from the front page (72) Inventor Yukiyasu Ito 5030 Totsukacho, Totsuka-ku, Yokohama-shi, Kanagawa Prefecture Software Development Division, Hitachi, Ltd.

Claims (11)

[Claims] 1. A set of records including one or more items is set as target data, one or more of the items are set as category items, and the set of records is classified into a plurality of categories according to the values of the category items. Category classification means; category evaluation means for calculating a statistic for each of the plurality of categories by a predetermined algorithm, and calculating an evaluation value for each of the plurality of categories based on the statistic; A data analysis system comprising: a category selection unit that selects one or a plurality of categories from the plurality of categories based on a value; and a presentation unit that presents the selected category to a user. 2. The data analysis system according to claim 1, further comprising: means for displaying statistics of one or more selected categories in a table format. 3. The data analysis system according to claim 1, further comprising: means for setting one or more reference categories by a user, wherein said category classification means sets said set of records by a value of said category item. A data analysis system that classifies the data into a plurality of categories included in the reference category. 4. The data analysis system according to claim 1, further comprising: means for setting one or more categories selected by the category selecting means as a new reference category. A data analysis system characterized by repeatedly executing processing of setting a category, classifying into a category, calculating an evaluation value, selecting a category, and setting a new reference category. 5. The data analysis system according to claim 1, further comprising: means for storing a set of explanation items about category items and statistics, wherein one or more selected by the category selection means. Means for calculating a statistic of the selected one or more categories from the target data based on the value of the category item for the category, and a fragment of the description sentence according to the statistic and the value of the category item And a means for synthesizing a description explaining the characteristics of the statistic of the selected one or more categories, and a means for displaying the description. 6. The data analysis system according to claim 2, further comprising holding means for holding a value of a category item and an evaluation value of the category for one or more selected categories, and displaying the statistics in a table format. Means for displaying, when displaying the held content of the holding means, means for highlighting a table element corresponding to a category having a maximum or minimum evaluation value among the selected one or more categories. Characteristic data analysis system. 7. The data analysis system according to claim 4, wherein said one or more categories sequentially selected by said category selection means in said iterative execution processing maintain and maintain the order of selection. Data, comprising: a selection result holding unit; and a unit that displays, in a list format, the contents held by the category selection result holding unit in accordance with the order selected during the execution of the repetitive execution process or after the end of the repetition execution process. Analysis system. 8. The data analysis system according to claim 4, wherein, in the iterative execution processing, a statistic of one or more categories selected at a certain time is calculated by:
A data analysis system comprising: means for presenting to a user during execution of the repetitive execution processing, and simultaneously presenting one or more categories selected at the next time to the user. 9. The data analysis system according to claim 4, wherein said one or more categories sequentially selected by said category selection means in said repetitive execution processing are maintained while maintaining the order of selection. A data analysis system comprising: a selection result holding unit; and a unit for presenting contents held in the category selection result holding unit to a user after the repetition execution process is completed in a selected order. 10. A system comprising: a first device and a second device connected by a data transfer path, wherein the first device generates a plurality of categories based on item information relating to target data; Evaluation value calculating means for calculating an evaluation value of each of the plurality of categories based on the category information and the statistics of each of the plurality of categories, and selecting one or more of the plurality of categories based on the evaluation value Category selection means to be
Presenting means for presenting the selected category to a user, the second device comprising: a storage means for storing a set of records including one or more items as target data; And a statistic holding means for holding the statistic calculated by the means, wherein the statistic calculation means of the second device comprises: Calculating the statistic based on the category information transferred from the device via the data transfer path and the target data in the storage means, and transferring the calculated statistic to the first device via the data transfer path The evaluation value calculation means of the first device calculates the evaluation value based on a statistic transferred from the second device via a data transfer path. Beam. 11. A storage medium storing a processing procedure in the data analysis device according to claim 1 as a computer program.