JP3004067B2 - Record classification method for information processing device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for classifying records in an information processing apparatus, and more particularly, to a method suitably applied to a method utilizing discriminant analysis processing.

[0002]

2. Description of the Related Art As an information processing apparatus, there is an information processing apparatus which performs information processing in a so-called record unit in which a plurality of types of data (so-called fields) are grouped. In this case, a plurality of records may be classified from a predetermined viewpoint, and processing may be performed for each classification or for any of the classifications. For example, records consisting of fields such as branch name, branch location, sales floor area, sales, profit, and number of employees are classified based on sales, and there is a difference between groups classified as profit per employee. It is also performed to detect whether or not the operation is performed. In this example, sales are in the taxonomy field.
ing. Here, when a new record is input, it is determined to which group the record belongs. In this case, if there is data in the classification scale field that defines the group to be classified, classification may be performed based on the value. However, if the data of the classification scale field is unknown, the classification group is determined from the data of the other fields. Is determined (classified). Discrimination analysis processing, which is a method of multivariate analysis, is often used to determine such belonging groups. The discriminant analysis process is also used for determining the validity of selecting a classification scale and the validity of classification conditions.

[0003] The concept of discriminant analysis processing (strictly speaking, educational discriminant analysis processing) will be briefly described with reference to FIG. FIG. 2 shows an example in which belonging groups G1 and G2 are determined from two field data x1 and x2.

[0004] A discriminant function D, which is a plane defining a boundary between groups, is determined in advance from records R11, R12, ... and R21, R22, ... in which the belonging groups G1 or G2 are known.
As is apparent from FIG. 2, the discriminant function D is represented by a straight line in the two-dimensional coordinate system, so that each field data x1
, And x2, which is the projection plane of the coordinates of each record that maximizes the difference between groups and minimizes variations within the groups.

The records R11, R12,... And R21,
The determination of the belonging group G1 or G2 of R22,... Depends on the data (x3) of the classification scale field different from the above-mentioned field data x1 and x2.

Now, it is assumed that a record Rx whose data value x3 of the classification scale field is unknown is input. At this time, a position in the coordinate system is obtained from the field data x1 and x2 of the record Rx, and based on the position DRx projected on the surface defined by the discriminant function D, the group (G1) to which the record Rx belongs is determined. decide.

Accordingly, a conventional series of processes for determining the group to which a record whose affiliation is unknown belongs is as shown in FIG. First, the discriminant analysis processing is started, and then, for example, a partial record is selected from records already stored in the auxiliary storage device or the like, and the designation of the group to which the selected record belongs is taken in (steps 101 to 103). . After that, various calculations are executed to determine a discriminant function and the like, and the processing result of the obtained discriminant function and the like is displayed and recognized by the user (step 1).
04, 105). Upon completion of such processing, a record of unknown affiliation is acquired to determine an affiliation group, the result is displayed in, for example, a table format, and a series of processing is terminated (10).
6, 107).

As described above, in order to obtain the discriminant function,
It is necessary to select and specify a plurality of sets of records whose affiliations are already known. As a specific method, a separate data file is formed for each group based on an existing data file, and group identification information is added to a record.

[0009]

However, in the prior art, when specifying an affiliation group of a record to be input in order to calculate a discriminant function or the like, the user is required to classify the record regardless of which specific method is adopted. The data value of the scale field is determined for each record, and the belonging group is designated one by one. Therefore, it requires much time and effort.

[0010] In order to improve the classification accuracy of records, a large number of input records whose affiliations are specified are required. Therefore, the above-mentioned problem of performing the specifying operation for each record is significant.

The present invention has been made in view of the above points, and is an information processing apparatus capable of easily specifying an affiliation group for an input record in discriminant analysis processing and shortening a series of processing time. It is an object of the present invention to provide a method for classifying records of an apparatus.

[0012]

In order to solve this problem, according to the present invention, a record whose affiliation group is unknown is classified into any group, and each record whose affiliation group is designated is assigned to the affiliation group. record classification method smell <br/> of an information processing apparatus for performing discriminant analysis process or to calculate the probability of belonging Te, the designation of affiliation group of each record in the input data for the discriminant analysis processing, the following processing I decided to do it.

That is, a plurality of files constituting a record
Selection process to select one or more of the fields as the classification scale field for the classification scale, and conditions for applying to the data of the selected classification scale field to determine which group each record belongs to. The affiliation group of each record, which is input data for the discriminant analysis processing , is specified by the fetching condition import process and the determination process of determining the affiliation group of each record based on the fetched conditions.

[0014]

In the information processing apparatus, discriminant analysis is used in the process of classifying records whose affiliation group is unknown into any group or calculating the probability that each record whose affiliation group is designated belongs to the affiliation group. Processing is used. When performing the discriminant analysis process, it is necessary that the group to which each record as input data belongs belongs.

The present invention relates to the designation of the group to which such an input record belongs. In the present invention, first, a record is constructed
A plurality of fields of any one or more of, is selected as the classification metric field defining the belonging group, then each applied to the data of the classification scale field selected records
Then, a condition for determining which group the record belongs to is fetched, and finally, the group to which each record belongs is determined based on the fetched condition. After the determination of the belonging group, the process immediately proceeds to the discriminant analysis process.

[0016]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below in detail with reference to the drawings. The information processing apparatus according to this embodiment, like a general computer system, includes a CPU, a memory, a keyboard, a CRT display, and the like in terms of hardware, although not shown. The information processing apparatus of this embodiment is configured as shown in FIG. 4 from the function of record classification. Note that the functional configuration shown in FIG. 4 is actually configured by software.

As shown in FIG. 4, the functional configuration (program structure concept) is divided into a control module 11, an input processing module 12, an output processing module 13, a table management processing module 14, an emphasis processing module 15, and a statistical processing module 16. Can be

The control module 11 is responsible for overall control, and the various modules 12 to 16 execute predetermined processing under the control of the control module 11.

The input processing module 12 captures records and commands input from an input device such as a keyboard. For example, a record having data of a classification scale field for determining a classification or a record whose data of a classification scale field is unknown and whose affiliation group is unknown is taken in. It also accepts designation (selection) of a classification scale field and incorporates conditions for determining a group of input records.

The output processing module 13 controls output to a CRT display, a printer, and the like. For example, it outputs an image to be used for specifying the group to which the input record belongs, outputs the obtained discriminant function, and outputs the group information obtained for a record whose belonging is unknown.

The table management processing module 14 manages table data including a plurality of records stored in an auxiliary recording device, a main memory or the like. In the case of this embodiment, the table data managed by the module 14 may be provided with emphasis information indicating that a part of a plurality of records is emphasized more than other records.

The emphasis processing module 15 is for emphasizing some records in the table data managed by the table management processing module 14 that satisfy a predetermined condition while distinguishing them from other records. That is, the record group is divided into an emphasized record group and a non-emphasized record group, and the belonging group of the input record necessary for determining the discriminant function or the like is designated. The emphasis processing module 15
More specifically, it includes a condition definition unit 15a, a determination unit 15b, and an emphasis processing unit 15c. The condition definition unit 15a activates the input processing module 12, and takes in the designation (selection) of the classification scale field and the condition for determining whether or not the record is to be emphasized based on the data value of the classification scale field. The determination unit 15b determines whether each record should be emphasized according to the conditions. Emphasis processing unit 1
5c is a table management processing module 14 based on the determination result.
Is to add emphasis information to a table under the management of the above.

The statistical processing module 16 calculates a discriminant function or the like and determines the affiliation destination of an affiliation unknown record. The statistical processing module 16 includes a discriminant analysis unit 16a and a probability calculation unit 16b. The discriminant analysis unit 16a calculates information such as a discriminant function using both the emphasized record group and the unemphasized record group.
The probability calculation unit 16b, when a record whose affiliation is unknown is given, determines the affiliation group of the record using the discriminant function and other information calculated by the discriminant analysis unit 16a. Is represented by the probability of belonging to one of the belonging groups. Note that the probability calculation unit 16b also calculates the belonging probability of an input record for which a belonging group is specified.

Next, a series of record classification processing for determining a group to which a record whose affiliation is unknown (classifying records) will be described with reference to FIGS. 1, 5 to 7. FIG. FIG. 1 is a flowchart showing the flow of the processing. 5 and 6 are explanatory diagrams showing table data as input records used for determining a discriminant function and the like. FIG. 5 shows a table before the emphasis processing, and FIG. 6 shows a table after the emphasis processing. . FIG. 7 is an explanatory diagram showing the result of determining the group to which the unknown record belongs. The example shown in FIG. 5 to FIG.
One record is formed by four fields A, B, C, and D.

First, the classification scale field selected by the user is fetched (step 201). In this case, a plurality of fields may be designated as classification scale fields. Next, conditions for the classification scale field for allocating each input record to two groups are fetched (step 202). When a plurality of fields are designated as the classification scale fields, the condition for each field and the overall condition for the result are taken. And
A determination is made based on the fetched conditions, and information indicating whether the record is an emphasized record or a non-emphasized record is added to the record (step 203).

FIG. 6 shows an emphasis result when field A is selected as the classification measure field and "value of field A>600" is specified as the condition of the emphasis record.

Thereafter, the records are automatically divided into two groups based on the emphasis result without waiting for designation by the user (step 204). In the case of the example in FIG. 6, the record group is divided into record groups with record numbers 1, 2, 4, 5, and 6, and record groups with record numbers 3, 7, 8, 9, and 10.

In this state, the discriminant analysis process is started, and each of the sorted record groups is automatically designated as an input record group used for discriminant analysis (step 2).
05, 206). Thereafter, a discriminant analysis process is executed to determine a discriminant function and the like, and the processing result of the obtained discriminant function and the like is displayed and recognized by the user (steps 207 and 207).
8). When the preprocessing for such classification is completed, a record whose affiliation is unknown is fetched, an affiliation group is determined based on a discriminant function or the like (for example, an affiliation probability is calculated), and the result is expressed in a table format or the like. Is displayed and a series of processing ends (20
9, 210).

FIG. 7 shows two records of unknown affiliation whose values of the classification scale field A are unknown (record numbers 11 and 1).
2) shows the result when given, and in this case, these records are both groups on the highlighting side (field A
Belonging to the group whose value is larger than 600) (emphasis probability)
Is larger than the probability of belonging to the group on the non-emphasized side.

FIG. 7 also shows the probabilities of input records for which a group is specified. For example, although the record with the record number 6 is specified to belong to the non-emphasized group, the probability of belonging to the emphasized group is considerably high from the result of the discriminant analysis. Such results are also used for selecting a classification scale field and analyzing the validity of classification conditions.

Therefore, according to the above-described embodiment, in order to execute a discriminant analysis process for classifying a record whose affiliation is unknown or calculating a probability that a record whose affiliation is designated belongs to the affiliation group. Is automatically performed according to the selection of the classification scale field and the setting of the conditions, so that the processing time and the labor of the user can be significantly reduced. In other words, there is no need to perform input operations after allocating records to a plurality of groups, input without being aware of the groups, and then specify the classification scale field and set conditions, which improves the usability of the user. I do. Also, already,
Even when using records stored in the storage device,
The belonging group can be easily specified.

In practice, an information processing device such as a database device or a spreadsheet device processes a large number of records,
It is necessary to classify records as necessary and judge the validity of the classification. Since the records are often used as in the above-described embodiment, the above-described effects are significant.

Incidentally, some conventional information processing apparatuses such as a database apparatus and a spreadsheet apparatus have a process of emphasizing some records so as to be different from other records. It is a process provided separately from the analysis process, and therefore cannot be used directly in the discriminant analysis process.
After executing the emphasis processing, the user needs to perform processing so that the processing can be used in the discriminant analysis processing, which is also time-consuming and troublesome as compared with the embodiment.

In the above-described embodiment, the discrimination into two groups is described. However, the invention can be applied to the discrimination (determination) into three or more groups. Also, one record consists of four field data, but the present invention is not limited to this. Furthermore, the present invention can be applied to discriminant analysis processing in which classification processing is not performed on records whose affiliation group is unknown.
That is, the record classification referred to in the claims includes calculating the belonging probability for the record to which the belonging group is designated.

[0035]

As described above, according to the present invention, the specification of a plurality of groups of input records, which is required in the discriminant analysis processing for performing record classification, is performed by selecting the classification scale field and the conditions for the field data. , The operation by the user can be greatly reduced, and the processing time, including the specification of the input record, can be made much shorter than before.

[Brief description of the drawings]

FIG. 1 is a processing flowchart of an embodiment method .

FIG. 2 is a schematic explanatory diagram of a discriminant analysis process.

FIG. 3 is a processing flowchart of a conventional method .

FIG. 4 is a functional block diagram of the embodiment method .

FIG. 5 is an explanatory diagram of table data before emphasis processing.

FIG. 6 is an explanatory diagram of table data after emphasis processing.

FIG. 7 is an explanatory diagram of a final processing result.

[Explanation of symbols]

11: control module, 12: input processing module, 1
3 output processing module, 14 table management processing module, 15 emphasis processing module, 16 statistical processing module.

──────────────────────────────────────────────────続 き Continuation of the front page (56) References JP-A-1-288961 (JP, A) JP-A-2-24779 (JP, A) “Program product VOS3 Statistical calculation library HISTATE 2 Analysis program compilation grammar” , P p. 277-291 (March 1990) (58) Fields investigated (Int. Cl. ⁷ , DB name) G06F 17/30 G06F 12/00 510

Claims (1)

(57) [Claims] An information processing apparatus for performing a discriminant analysis process of classifying records whose affiliation group is unknown into any group, and calculating a probability that each record whose affiliation group is specified belongs to the affiliation group. Record classification method , any one or more of a plurality of fields constituting a record
A selection of as the classification scale field for the classification scale
A selection process for a taking process conditions capture the conditions which determine whether each record by applying belongs to which group the data classification scale field selected, affiliation of each record on the basis of the captured conditions A record classification method for an information processing apparatus, comprising: specifying a group to which each record, which is input data for a discriminant analysis process , is assigned by a discriminating process for determining a group, and then proceeding to a discriminant analysis process.