JP7197391B2 - Data analysis device, data analysis method and data analysis program - Google Patents

Description

The present invention relates to a data analysis device, data analysis method and data analysis program.

Patent Literature 1 discloses an advertisement creation support device that assists creation of advertisements based on analysis of the relationship between product features and past sales performance (see paragraph 0006). Predetermined product attributes are set as variables, and past sales results of products are set as objective variables (see paragraph 0007).

JP 2010-237923 A

Thus, conventionally, multiple objective variables could not be handled at the same time when inferring some rule in which the value of the objective variable is determined by the value of the explanatory variable. For example, in Patent Literature 1, only "past sales performance of products" is set as an objective variable.

For this reason, conventionally, in order to estimate the values of a plurality of objective variables, the number of combinations of explanatory variables and objective variables becomes enormous, and there is the problem that it is difficult to understand the analysis results. In other words, conventionally, the analysis results must be created for the number of objective variables included in the data to be analyzed, and the analysis results are enormous in volume and difficult to understand.

The present invention has been made in view of the above problems, and is a data analysis apparatus capable of performing data analysis in a manner that can collectively output as one analysis result even when data to be analyzed includes a plurality of target variables. , to provide a data analysis method and a data analysis program.

In order to solve the above-described problems and achieve the object, a data analysis device according to the present invention is a data analysis device comprising a control unit, the control unit having a record having an explanatory variable and a plurality of objective variables. A record dividing means for dividing the records in the analysis target data including a plurality of ratio calculating means for calculating, for each of the record groups, a mode value and a ratio of the mode value within the same objective variable for all of the plurality of objective variables; analysis result creating means for creating an analysis result including the mode value and the ratio calculated by the ratio calculating means for all of the plurality of objective variables for each value of the explanatory variable as and

Further, in the data analysis apparatus according to the present invention, the record dividing means repeats the dividing using the explanatory variable other than the explanatory variable used as the dividing reference as a new reference, and the record group obtained by the dividing according to at least one of the number of records in the table, the number of times the division is executed, and the impurity improvement amount, which is the improvement amount of the impurity for the objective variable between before the division and after the division. and stopping the division.

Further, in the data analysis device according to the present invention, the improvement amount calculated for each objective variable by the control unit as the impurity improvement amount is a value obtained by subtracting the impurity after the division from the impurity before the division. a calculating means for outputting only the mode value and the ratio after the division for the objective variable for which the impurity improvement amount calculated by the improvement amount calculating means is equal to or greater than a predetermined value among the analysis results; and analysis result output means for outputting the results.

Further, in the data analysis device according to the present invention, the control unit subtracts the sum of the squares of the ratios of the values in the same objective variable from 1, and calculates the impurity as the impurity It is characterized by further comprising calculating means.

Further, in the data analysis apparatus according to the present invention, the analysis result creation means hierarchically branches the analysis results based on the explanatory variables used as the criteria for the division by the record division means and the values of the explanatory variables. creating a decision tree including the mode and the ratio calculated by the ratio calculating means for all of the plurality of objective variables, for each branch that is set.

Further, in the data analysis device according to the present invention, the explanatory variable is an item for specifying the status of an employee, and the objective variable is for determining a method of calculating or paying a salary to the employee. It is characterized by being an item of

Further, a data analysis method according to the present invention is a data analysis method executed by an information processing apparatus having a control unit, wherein a plurality of records having an explanatory variable and a plurality of objective variables are executed by the control unit. a record splitting step of splitting the records in the analysis target data including the A ratio calculation step of calculating, for each record group, a mode value and a ratio of the mode value within the same target variable for all of the plurality of target variables, and the record splitting step set the splitting criteria as the splitting criteria. and an analysis result creation step of creating an analysis result including the mode and the ratio calculated in the ratio calculation step for all of the plurality of objective variables for each value of the explanatory variable. .

Further, a data analysis program according to the present invention is a data analysis program to be executed by an information processing apparatus having a control unit, the record having an explanatory variable and a plurality of objective variables to be executed by the control unit A record dividing step of dividing the records in the data to be analyzed, which includes a plurality of a ratio calculating step of calculating, for each of the plurality of objective variables, a mode value and a ratio of the mode value within the same objective variable for each record group; and a dividing criterion in the record dividing step. and an analysis result creation step of creating an analysis result including the mode and the ratio calculated in the ratio calculation step for all of the plurality of objective variables for each value of the explanatory variable as and

ADVANTAGE OF THE INVENTION According to this invention, it is effective in the data analysis in the aspect which can output collectively as one analysis result, even when the data used as an analysis object contain several target variables.

FIG. 1 is a block diagram showing an example of the configuration of an information processing device. FIG. 2 is a diagram illustrating an example of analysis target data. FIG. 3 is a diagram showing an example of impurity calculation. FIG. 4 is a diagram showing an example of calculation of the total impurity improvement amount when the data to be analyzed is divided on the basis of X1. FIG. 5 is a diagram showing an example of calculation of the total impurity improvement amount when the data to be analyzed is divided on the basis of X2. FIG. 6 is a diagram showing an example of calculation of the total impurity improvement amount when the data to be analyzed is divided based on X3. FIG. 7 is a diagram showing an example of data after division when analysis target data is divided based on X3. FIG. 8 is a diagram illustrating an example of data after division when data division is continued. FIG. 9 is a diagram showing a part of an example of data after division when the data is further divided. FIG. 10 is a diagram showing an example of calculation of the amount of improvement in impurity before and after division (influence score of division) in each division. FIG. 11 is a diagram showing an example of outputting analysis results for each condition (by explanatory variable). FIG. 12 is a diagram showing an example of outputting analysis results for each condition (for each explanatory variable) in consideration of the division influence score. FIG. 13 is a diagram showing an example of outputting analysis results in the form of a tree (decision tree). FIG. 14 is a diagram showing an example of outputting analysis results in the form of a tree (decision tree) in consideration of the influence score of division or the impurity after division. FIG. 15 is an image diagram showing an example of setting employee information when an employee joins the company. FIG. 16 is an image diagram showing an example of setting employee information when an employee joins the company. FIG. 17 is an image diagram showing an example of registration of a plurality of items (50 items) for a plurality of new employees (50 persons). FIG. 18 is an image diagram showing an example of occurrence of incorrect payment of salary due to omission of registration of employee information for a new employee. FIG. 19 is a diagram showing an example of the contents of employee data. FIG. 20 is a diagram showing an example of the contents of each master. FIG. 21 is a diagram showing an overview of the flow of processing up to template registration (registration of definition master contents). FIG. 22 is a diagram showing an overview of the flow of processing from template registration (registration of definition master contents) to employee information registration. FIG. 23 is a diagram showing an example of an analysis execution screen. FIG. 24 shows the display of registration candidates for the selected branch (indicated by MA4), the display of item-by-item adjustment (indicated by MA5), and the display of the final registration candidate in response to the selection result of item-by-item adjustment (indicated by MA3). ) is performed on the pattern registration screen. FIG. 25 is a diagram showing an example of setting a definition name (indicated by MB1) and registering a pattern (indicated by MB2) for a pattern displayed as a final registration candidate on the pattern registration screen. FIG. 26 is a diagram showing an example of confirming and editing registered pattern information on the employee information registration/change screen. FIG. 27 is a diagram showing an example of how to call patterns from the employee information registration/change screen and how to display the called patterns on another screen. FIG. 28 is a diagram showing an example of an employee information registration/change screen in which employee information is automatically set by selecting a specific pattern from the called patterns. FIG. 29 is an image diagram showing an example of setting employee information. FIG. 30 is an image diagram showing an example of changing employee information. FIG. 31 is an image diagram showing an example of the work to be done by the person in charge of payroll when changing employee information. FIG. 32 is an image diagram showing an example of occurrence of incorrect payment of salary due to failure to change employee information about an existing employee. FIG. 33 is a diagram showing an overview of the flow of processing from template registration (registration of definition master contents) to change of employee information. FIG. 34 shows the display of registration candidates for the selected branch (indicated by MF4), the display of item-by-item adjustment (indicated by MF5), and the display of the final registration candidate in response to the selection result of item-by-item adjustment (indicated by MF3). ) is performed on the pattern registration screen. FIG. 35 is a diagram showing an example of setting a definition name (indicated by MG1) and registering a pattern (indicated by MG2) for a pattern displayed as a final registration candidate on the pattern registration screen. FIG. 36 is a diagram showing an example of confirming and editing registered pattern information on the definition information registration/change screen. FIG. 37 is a diagram showing an example of changing an employee classification from the employee information registration/change screen. FIG. 38 is a diagram showing an example of a registered pattern list displayed based on the changed employee classification and a definition information registration/change screen displayed for confirming the contents of the pattern. FIG. 39 is a diagram showing an example of an employee information registration/change screen in which employee information is automatically changed by selecting a specific pattern from the displayed pattern list.

Embodiments of a data analysis device, a data analysis method, and a data analysis program will be described in detail below with reference to the drawings. It should be noted that the present invention is not limited by this embodiment.

Specific configurations and operations will be described below, but the contents of the data analysis device, data analysis method, and data analysis program are mainly described in [1-1], [2-1], and [3-1] below. ] and [4-1].

[1. composition]
An example of the configuration of an information processing apparatus 100 including the present invention will be described with reference to FIG. FIG. 1 is a block diagram showing an example of the configuration of an information processing apparatus 100. As shown in FIG.

The information processing device 100 is a commercially available desktop personal computer. Note that the information processing apparatus 100 is not limited to a stationary information processing apparatus such as a desktop personal computer, and may be a portable type such as a commercially available notebook personal computer, PDA (Personal Digital Assistants), a smart phone, or a tablet personal computer. It may be an information processing device.

The information processing apparatus 100 includes a control section 102 , a communication interface section 104 , a storage section 106 and an input/output interface section 108 . Each unit included in the information processing apparatus 100 is communicably connected via an arbitrary communication path.

The communication interface unit 104 communicably connects the information processing device 100 to the network 300 via a communication device such as a router and a wired or wireless communication line such as a dedicated line. The communication interface unit 104 has a function of communicating data with another device via a communication line. Here, the network 300 has a function of connecting the information processing apparatus 100 and the server 200 so that they can communicate with each other, and is, for example, the Internet or a LAN (Local Area Network). Data such as various masters described later may be stored in the server 200, for example.

An input device 112 and an output device 114 are connected to the input/output interface section 108 . The output device 114 can be a monitor (including a home television), a speaker, or a printer. The input device 112 can be a keyboard, a mouse, a microphone, or a monitor that realizes a pointing device function in cooperation with a mouse. Note that, hereinafter, the output device 114 may be referred to as the monitor 114 and the input device 112 may be referred to as the keyboard 112 or the mouse 112 .

The storage unit 106 stores various databases, tables, files, and the like. The storage unit 106 stores a computer program for performing various processes by giving commands to a CPU (Central Processing Unit) in cooperation with an OS (Operating System). As the storage unit 106, for example, memory devices such as RAM (Random Access Memory) and ROM (Read Only Memory), fixed disk devices such as hard disks, flexible disks, and optical disks can be used.

The storage unit 106 stores, for example, analysis target data 106a, an automatically updated definition master 106b as a definition master (hereinafter simply referred to as "definition master 106b"), an employee classification master 106c, a salary classification master 106d, a position master 106e, grade master 106f, occupation master 106g, work location master 106h, and calculation definition master 106i.

As shown in FIG. 2, the analysis target data 106a includes multiple records having explanatory variables (X1, X2 and X3) and multiple objective variables (Y1 and Y2). The explanatory variable is a variable that explains the objective variable, and can be regarded as the cause of things. The objective variable is a variable to be predicted, and can also be regarded as a result of things. A major feature of the information processing apparatus 100 according to the present embodiment is that it can handle a plurality of objective variables (Y1 and Y2).

The analysis target data 106a corresponds to, for example, the employee data 106a1 in the case of salary payment.

When the analysis target data 106a is the employee data 106a1, the explanatory variable corresponds to the status item that is an item for specifying the status of the employee, and the value of the explanatory variable is the value of the status item. equivalent to the value Examples of the status items include employee classification, gender, title, work location, presence or absence of qualifications, duties, and the like. As the status item values, for example, the status item value of the status item "employee classification" is "regular employee", the status item value of the status item "gender" is "male", and the status item is "work". "Sales registrant", etc., which is the identity item value of .

When the analysis target data 106a is the employee data 106a1, the objective variable corresponds to a salary item which is an item for determining how to calculate or pay the employee's salary, and the value of the objective variable is , corresponds to the salary item value, which is the value of the salary item. The salary items include, for example, salary category, salary payment date category, working condition category, insurance type, and presence/absence of various allowances. As the salary item value, for example, the salary item value of the salary item "salary classification" is "2: Daily wage earner" and the salary item value of the salary item "salary payment date classification" is "1: 25-day payment” and the like.

The contents of the employee data 106a1 are illustrated in FIG. 19. In FIG. 19, only some of the numerous explanatory variables and objective variables are shown. In the employee data 106a1, data of each employee is held for each salary payment date.

The definition master 106b is a master for defining patterns composed of the above-mentioned status items and above-mentioned status item values, and above-mentioned salary items and above-mentioned salary item values. The definition master 106b includes, for example, the pattern and pattern definition data (definition name, definition number, etc.) for defining the pattern in association with each other.

The contents of the employee division master 106c, salary division master 106d, position master 106e, grade master 106f, occupation master 106g, work location master 106h, and calculation definition master 106i are, for example, as shown in FIG. The content of the employee classification master 106c does not differ much depending on the company. The content of the grade master 106f may be set completely differently depending on the company. At the time of payroll calculation, the "update conditions" in the calculation definition master 106i corresponding to the "job type" and "adjusted pay category" of the employee data 106a1 are acquired, and the calculation definition master 106i associated with the acquired "update conditions" is updated. Payroll can be calculated based on the "renewal formula".

The control unit 102 is a CPU or the like that comprehensively controls the information processing apparatus 100 . The control unit 102 has an internal memory for storing a control program such as an OS, a program defining various processing procedures, required data, and the like, and performs various information processing based on these stored programs. Run.

Functionally, conceptually, for example, the control unit 102 (1) compares the records in the analysis target data including a plurality of records having an explanatory variable and a plurality of objective variables with the specific explanatory variable as a reference, (2) a record dividing unit 102a as a record dividing unit that divides into a record group for each value of the explanatory variable; (3) a ratio calculation unit 102b as ratio calculation means for calculating the ratio of the frequent value and the mode value for all of the plurality of objective variables; (4) an analysis result creation unit 102c as analysis result creation means for creating an analysis result including the mode value and the ratio calculated by the ratio calculation means for each of the plurality of objective variables; (5) the analysis result; Among them, as analysis result output means for outputting only the mode value and the ratio after the division for the objective variable for which the impurity improvement amount calculated by the improvement amount calculation means is equal to or greater than a predetermined value an analysis result output unit 102e; a calculation unit 102f; and (7) determining, within the same salary item, the set of employees specified by a branch selected from branches hierarchically branched based on the status item and the value of the status item. The salary item value with the largest ratio is displayed on a pattern registration screen, which is a screen for performing the registration of the pattern, together with the largest ratio, and when the largest ratio is equal to or greater than a predetermined threshold, , a registration candidate pattern as a registration candidate pattern display means for displaying the salary item value of the ratio of the threshold value or more on the pattern registration screen in a manner that can be identified as the registration candidate, together with the ratio of the threshold value or more. a display unit 102g; The pattern consisting of salary items and the pattern for defining the pattern (9) the identification item values for each employee from the definition master; and acquiring the item and the value for the pattern associated with the pattern definition data specified on the employee information registration/change screen, which is a screen for newly registering or changing the salary item value, and acquiring the acquired item (10) a registration candidate item display unit 102i as registration candidate item display means for displaying the value of the above as a candidate for new registration on the employee information registration/change screen; When the identity item value for a specific identity item is changed on the employee information registration/change screen, which is a screen for newly registering or changing the salary item value, the change is made from the definition master. (11) from the definition master, a pattern definition display unit 102j as pattern definition display means for acquiring pattern definition data associated with the identification item and the changed identification item value, and displaying the acquired pattern definition data; , acquiring the item and the value of the pattern associated with the specified one of the pattern definition data displayed by the pattern definition display means, and displaying the value of the acquired item as the change candidate; and a change candidate item display unit 102k as change candidate item display means for displaying on the employee information registration/change screen. The operations performed by each unit will be described below.

[1-1. Operation for Creation of Decision Tree]
The record dividing unit 102a divides the records in the analysis target data 106a, which includes a plurality of records having an explanatory variable and a plurality of objective variables, into records for each value of the explanatory variable with reference to the specific explanatory variable. Divide into groups. For example, as shown in FIG. 8, the record dividing unit 102a divides 10 records in the analysis target data 106a into a record group A consisting of two records having X3=1 with reference to the explanatory variable X3, and X3 A record group D consisting of 6 records with X3=2 and a record group I consisting of 2 records with X3=3.

The record dividing unit 102a repeats the division using the explanatory variable other than the explanatory variable used as the division criterion as a new criterion, and executes the division based on the number of records in the record group obtained by the division. The division may be stopped according to at least one of the number of times the division is performed and an impurity improvement amount that is an improvement amount of the impurity for the objective variable between before the division and after the division. For example, as shown in FIG. 8, the record dividing unit 102a divides two records in the record group A into a record group B consisting of one record having X2=1 and one record having X2=2. 6 records in the record group D are divided into a record group E consisting of 3 records having X1=1 and a record group consisting of 2 records having X1=2 Divide into F and . Further, as shown in FIG. 8, the record dividing unit 102a divides the three records in the record group E into a record group G consisting of two records having X2=1 and one record having X2=2. and a record group H consisting of records of . In this manner, the record dividing unit 102a divides ten records in the analysis target data 106a into record groups A to I, as shown in FIG.

A ratio calculation unit 102b calculates, for each of the record groups obtained by dividing by the record division unit 102a, the mode value in the same objective variable and the ratio of the mode value to all of the plurality of objective variables. do. For example, when focusing on the record group A in FIG. 8, the ratio calculation unit 102b calculates that the mode value of the objective variable Y1 is 1, and the ratio of the mode value 1 is 1/2×100=50. For the objective variable Y2, the mode is calculated as 1, and the ratio of the mode 1 is calculated as 2/2×100=100%. Similarly, when focusing on the record group E in FIG. 8, the ratio calculation unit 102b calculates that the mode value of the objective variable Y1 is 1, and the ratio of the mode value 1 is 2/3×100≈ For the objective variable Y2, the mode is calculated as 2, and the ratio of the mode 2 is calculated as 3/3×100=100%.

The analysis result creation unit 102c performs an analysis including the mode value and the ratio calculated by the ratio calculation unit 102b for all of the plurality of objective variables for each value of the explanatory variable used as the basis for the division by the record division unit 102a. produce results. For example, when focusing on the record group A in FIG. 50%, the mode of the objective variable Y2 is 1, and the ratio is 100%." Similarly, when focusing on record group E in FIG. and its ratio is about 67%, the mode of the objective variable Y2 is 2, and its ratio is 100%." An image of the created analysis result is shown in FIG.

As the analysis result, the analysis result creation unit 102c calculates a ratio of all of the plurality of objective variables for each branch hierarchically branched based on the value of the explanatory variable used as the criterion for the division in the record division unit 102a. A decision tree including the mode and the ratio calculated in the unit 102b may be created. An image of the decision tree to be created is shown in FIG. In the analysis using a decision tree, for example, a variable to be analyzed in all data (objective variable) and a variable for branching data (explanatory variable) are set, the branching is repeated, and the analysis results are displayed in a tree structure. It is an analysis method expressed in a tree diagram.

The improvement amount calculation unit 102d calculates a value obtained by subtracting the impurity after the division from the impurity before the division as the impurity improvement amount for each objective variable. Here, the impurity can be calculated by the impurity calculation unit 102f by subtracting the sum of the squares of the ratios of the values in the same objective variable. For example, in FIG. 8, focus on the analysis target data 106a (=before division) and the record group A (=after division).

The impurity calculation unit 102f calculates the impurity of the objective variable Y1 in the analysis target data 106a (=before division) as 1-(percentage of Y1=1) ² -(percentage of Y1=2) ² =1-( 6/10) ² −(4/10) ² =0.48. Similarly, the impurity calculation unit 102f calculates the impurity of the objective variable Y2 in the analysis target data 106a (=before division) by 1−(percentage of Y2=1) ² −(percentage of Y2=2) ² − (Percentage of Y2 = 3) ² - (Percentage of Y2 = 4) ² = 1 - (2/10) ² - (3/10) ² - (3/10) ² - (2/10) ² = Calculate 0.53.

The impurity calculation unit 102f calculates the impurity of the objective variable Y1 in the record group A (=after division) as 1−(percentage of Y1=1) ² −(percentage of Y1=2) ² =1−(1 /2) ² - (1/2) ² = 0.5. Similarly, the impurity calculation unit 102f calculates the impurity of the objective variable Y2 in the record group A (=after division) as 1−(percentage of Y2=1) ² =1−(2/2) ² =0. calculate.

From the above, the improvement amount calculation unit 102d calculates the impurity improvement amount for the objective variable Y1 as follows: impurity 0.48 before division−impurity 0.5 after division=−0.02. Similarly, the improvement amount calculation unit 102d calculates the impurity improvement amount for the objective variable Y2 as follows: impurity 0.53 before division−impurity 0 after division=0.53.

Among the analysis results, the analysis result output unit 102e outputs the mode value and the ratio after the division for the objective variable for which the impurity improvement amount calculated by the improvement amount calculation unit 102d is equal to or greater than a predetermined value. can be output only. For example, if the predetermined value is assumed to be "0.05", the analysis result output unit 102e determines the mode value after the division for the objective variable Y1 whose impurity improvement amount is -0.02. and its ratio of 50% are not output, and only the mode of 1 after division and its ratio of 100% for the target variable Y2 whose impurity improvement amount is 0.53 are output. 12 and 14 show images of the display of analysis results in consideration of such an impurity improvement amount.

[1-2. Operations for registering and changing employee information]
The registration candidate pattern display unit 102g displays a branch selected from branches hierarchically branched based on the identification item and the identification item value (specifically, the left side of the pattern registration screen in FIGS. 24 and 34). "Conditional branch" branch of )), the salary item value that accounts for the largest percentage in the same salary item, together with the largest percentage, the registration of the pattern 24 and MF4 in FIG. 34). At this time, if the largest ratio is equal to or higher than a predetermined threshold, the registration candidate pattern display unit 102g displays the salary item values with a ratio equal to or higher than the threshold in a manner that can be identified as candidates for registration (specifically are hatched with oblique lines and dots) are displayed on the pattern registration screen (specifically, the area indicated by MA4 in FIG. 24 and MF4 in FIG. 34) together with the ratio of the threshold value or more.

The registration candidate pattern display unit 102g displays different values according to the numerical range to which the ratio equal to or higher than the threshold belongs (specifically, the salary item values belonging to the ratio of 100% are hatched with diagonal lines, and the ratio is 90%). Salary item values belonging to more than or equal to less than 100% are hatched with dots), and the salary item values of the percentage equal to or greater than the threshold value may be displayed.

The registration candidate pattern display unit 102g includes an adjustment button (specifically, MA2 in FIG. 24 and a When the "adjustment by item" button shown in MF2 of 34) is selected on the pattern registration screen (see FIGS. 24 and 34), a check box corresponding to the displayed salary item value is displayed and checked. The salary item values corresponding to the checked check boxes are displayed on the pattern registration screen (specifically, the "Select results" area).

The pattern registration unit 102h stores the identification item and the identification item value for the selected branch, the salary item value displayed in the identifiable manner by the registration candidate pattern display unit 102g, and the salary item corresponding thereto. (specifically, pattern 1 indicated by MA3 in FIG. 24 and pattern 2 indicated by MF3 in FIG. 34), and pattern definition data for defining the pattern, which is set by the operator (Specifically, the definition name corresponding to pattern 1 shown by MAB1 in FIG. (sales registrant)”) are linked and registered in the definition master 106b.

The registration candidate item display unit 102i displays an employee information registration/change screen (see FIG. 27, etc.) which is a screen for newly registering or changing the status item value and the salary item value for each employee from the definition master 106b. acquires the item and the value for the pattern associated with the pattern definition data specified in , and displays the value for the acquired item on the employee information registration/change screen as a candidate for new registration ( Specifically, it refers to the display of the area of ME2 in FIG. 28).

When the identification item value for a specific identification item is changed on the employee information registration/change screen (see FIG. 37, etc.), the pattern definition display unit 102j displays the above-mentioned The pattern definition data associated with the identity item and the changed identity item value are acquired, and the acquired pattern definition data is displayed (specifically, the display of the definition name on the upper screen of FIG. 38). The pattern definition display unit 102j may display the identification item and the identification item value for the pattern in the definition master 106b associated with the acquired pattern definition data (specifically, This refers to the display of condition items on the upper screen of FIG. 38).

The change candidate item display unit 102k acquires from the definition master 106b the item and the value for the pattern associated with the specified one of the pattern definition data displayed by the pattern definition display unit 102j, and The value of the item is displayed as a candidate for change on the employee information registration/change screen (specifically, it refers to the display of the area of MK1 in FIG. 39).

[2. Background and Overview]
In this section, we will explain the background and outline of creating a decision tree, registering employee information, and changing employee information.

[2-1. Background and overview of decision tree creation]
Conventional decision tree algorithms deal with only one value as an objective variable. In order to predict multiple values at the same time, it is necessary to analyze each of them, but in that case, the combination of conditional values and objective variables becomes enormous, and the problem is that it is difficult for users to understand the analysis results. . In addition, there is also a problem that it becomes difficult to understand which condition affects the result as the branching of conditions becomes deeper.

Therefore, in the present embodiment, for example, a plurality of objective variables can be handled simultaneously, thereby making it possible to present a user with a concise set of condition values and objective variable values. In addition, in this embodiment, for example, by scoring whether the condition is useful for determining the objective variable for each branch, it is possible to understand the rules between the condition and the objective variable more simply. As a result, even system users who do not specialize in analysis can utilize the analysis results.

Then, in this embodiment, for example, in order to find a rule between a plurality of values that can be a condition and a plurality of values that are considered to be determined by the conditions, in the data of the business system, there is a large amount of data. By analyzing the rules under which the data is entered, it is possible to predict other setting values based on a small number of items, reduce subsequent input errors, and improve input efficiency. enabled utilization. As input values, for example, discrete values such as segmented values are assumed. As a decision tree algorithm, CART, C4.5, etc. are known, but in this embodiment, by improving these known decision tree algorithms, it is possible to have many discrete value items like a business system. This makes it possible to apply the analysis to various data, making it easier to understand the analysis results.

[2-2. Background and overview of employee information registration]
Since each company has different salary regulations, the items and contents of "employee information" managed for payroll calculation vary from company to company. Here, in order to perform calculations in accordance with the company's salary regulations, it is necessary to set "employee information" on the system based on the salary regulations and contracts, but the contents written in the contract However, it was difficult to fully understand all the items and their contents that should be registered in the system by linking the contents described in the salary regulations from the beginning, and registration omissions and registration errors occurred. As a result, there were cases where erroneous payment occurred. In addition, the number of system items that need to be registered is large, and when there are many new employees joining the company at the same time, the amount of work is enormous.

Therefore, in this embodiment, for example, when registering "employee information", a large amount of existing data is analyzed, and at the same time, by providing a mechanism that can propose items to be input and their contents. , for example, to improve work efficiency and prevent erroneous payment. Specifically, in this embodiment, in order to solve the problem of wanting to know which items should be registered at the same time when an employee is registered, the information of other employees is analyzed and highly related items are confirmed. By doing so, we built a mechanism that allows us to grasp the items that should be registered at the same time. In addition, in this embodiment, in order to solve the problem of facilitating the input work when registering similar employees, the information of other employees is analyzed, and highly related items and their contents are identified. By selecting from the candidates, we built a mechanism that facilitates registration.

Below, in this item [2-2], the problems and background when registering employee information will be described in detail.

First, as shown in FIGS. 15 and 16, when an employee joins the company, the contents of salary regulations and employment contract are entered into the system. Regardless, input leaks may occur. For example, items that are not written in the employment contract, items whose input is determined by complicated conditions, and items that need to be set according to the company's own rules are prone to input omissions and input errors.

Figure 16 shows grade information and adjustment salary allowance (20,000 yen per month for employees who joined the company mid-career after 2010) as information that cannot be understood without referring to the salary regulations (usually thick sheets of paper, etc.). of information is exemplified. The person in charge of inputting the system needs to refer to the information on the adjustment salary allowance in the salary regulations and enter the adjustment payment category as "1: Target (a category that means that 20,000 yen will be paid)".

However, since the person in charge of inputting the system needs to input data into the system for many new employees, there is a problem that input omissions and input errors are likely to occur. For example, as shown in FIG. 17, when 50 items are entered into the system for each of 50 new employees, a total of 50 people×50 items=2500 items must be entered.

Due to such circumstances, if an input omission or an input error occurs, there is a possibility that it will lead to incorrect payment of salary. For example, as shown in Figure 18, if there is an omission in the input of the adjustment payment category "1: Eligible", as a result, the adjustment salary allowance of 20,000 yen will not be paid to those who were originally eligible for payment. As a result, there was a problem of erroneous payment.

[2-3. Background and overview of changes in employee information]
In order to calculate according to the company's salary regulations, it is necessary to set "employee information" on the system, but since the salary regulations differ for each company, it is necessary to manage " The items and contents of “employee information” vary from company to company. Here, since multiple items of "employee information" are complicatedly related to payroll calculation, it is often necessary to change many items at the same time when changing one item. However, in general, there is no description of the items that need to be changed on the system in the materials when it is necessary to change the "employee information". Therefore, when changing one item, in order to check the items that need to be changed at the same time, it is necessary to make a judgment by comparing the registered information of other employees, and the work was not efficient. In addition, there have been cases where erroneous payments have occurred as a result of failure to capture all the items that need to be changed and omissions of changes.

Therefore, in this embodiment, for example, when changing "employee information", a large amount of existing data is analyzed, and at the same time, a mechanism is provided that can propose items to be changed and their contents. , for example, to improve work efficiency and prevent erroneous payment. Specifically, in this embodiment, in order to solve the problem of wanting to know if there are other items to be changed when changing an item of employee information and to prevent omission of change, information of other employees is added. By analyzing this and confirming the items that are highly related to the content of the changes, we built a system that allows us to grasp other items that should be changed. In addition, in this embodiment, in order to solve the problem of wanting to easily register other necessary item changes when changing employee information items, information of other employees is analyzed and changed. We built a system that facilitates registration by selecting items with a high degree of relevance to the content and their content from candidates.

Below, in this item [2-3], the issues and background when changing employee information will be described in detail.

First, as shown in FIG. 29, when an employee joins the company or introduces the system, the employee information is set based on the salary regulations and the employment contract. Here, as shown in FIG. 30, when there is a change in information about an employee, the person in charge of payment of salaries, such as the personnel manager, sends a list of persons eligible for promotion, a list of persons eligible for affiliation transfer, and a list of persons eligible for re-employment after retirement. A list of eligible employees, a list of those eligible for change from part-time to regular employment, and a list of eligible employees for childcare/maternity leave, etc. will be sent. Here, from these lists, we can only know which employee is the target of the change. , employee E and employee F. For this reason, even if there is a need to change the "occupation" item of the three employees from general staff to managerial staff, this information is unknown.

Then, as shown in Fig. 31, if the "type" is a managerial position, a managerial allowance of 50,000 yen will be paid, and if the "type" is other than a managerial position, the managerial allowance will not be paid. If you forget to change the "type" of the above three employees from general staff to managerial staff on the system, as shown in Fig. 32, they are originally eligible for the management allowance. There was a problem that the allowance would not be paid to those (employee A, employee E, and employee F) who were previously paid, leading to erroneous payment.

[3. Details of processing]
In this section, a specific example of processing according to this embodiment will be described. The process of creating a decision tree, the process of registering employee information, and the process of changing employee information will be described in this order.

[3-1. Decision tree creation process]
First, decision tree creation processing will be described in detail with reference to FIGS. 2 to 14. FIG. In the decision tree algorithm of the present embodiment, the improvement amount for each division is calculated in (6) below, and the result of the calculation is used in (7) below to display results that are easy for the user to understand. One of the characteristics is that there is

(0) Data to be analyzed As a premise, it is assumed that the analysis target data 106a, which is data to be analyzed (=divided), is as shown in FIG. In the analysis target data 106a of FIG. 2, X1, X2 and X3 are explanatory variables, and Y1 and Y2 are objective variables. The explanatory variable and the objective variable are qualitative variables. In the following, it is assumed that there is some rule by which the value of the objective variable is determined by the value of the explanatory variable.

(1) Calculating Current Impurity First, the impurity calculation unit 102f calculates the impurity as an index representing the extent of data variation as follows. For impurity calculation, for example, information entropy, Gini coefficient, etc. used in a general decision tree algorithm can be used.

For example, if the impurity of Y1 in the analysis object data 106a of FIG. 2 is calculated using the Gini coefficient, the result is as follows. As shown in FIG. 3, the Genie coefficient is represented by a formula of 1-(percentage of Y1=1) ² -(percentage of Y1=2) ² . Referring to FIG. 2, the ratio of Y1=1 is 6/10, and the ratio of Y1=2 is 4/10. 10) Calculate ² - (4/10) ² = 0.48. From the above, the impurity calculation unit 102f calculates the impurity before division for Y1 as 0.48. Further, the impurity calculation unit 102f calculates the impurity of Y2 before division as 0.53 by the same calculation.

(2) For each condition item, calculate the amount of improvement in impurity when divided calculate.

(2-1) When dividing by X1 When dividing by X1, the record dividing unit 102a, as shown in FIG. A record group (X1=1 data) consisting of five records (id=a, b, c, d, e) having X1=2 and five records having X1=2 (id=f, g, h, i , j) (X1=2 data).

Here, the impurity calculation unit 102f calculates the impurity of Y1 in the data of X1=1 as 0.48 as shown in FIG. The Y2 impurity in the data is calculated as 0.48, the Y1 impurity in the X1=2 data is calculated as 0.48, and the Y2 impurity in the X1=2 data is calculated as 0.48.

Then, the impurity calculation unit 102f takes a weighted average of the impurities of Y1 calculated in the previous paragraph, and calculates the impurity after division for Y1 as 0.48. Similarly, the impurity calculation unit 102f takes the weighted average of the impurities of Y2 calculated in the previous paragraph, and calculates the impurity after division for Y2 as 0.48.

Finally, the improvement amount calculation unit 102d subtracts the impurity after the division of Y1 (0.48) from the impurity before the division of Y1 (0.48) to obtain the improvement amount of the impurity of Y1. is calculated as 0. Similarly, the improvement amount calculation unit 102d subtracts the impurity after the division of Y2 (0.48) from the impurity before the division of Y2 (0.53) to obtain the improvement amount of the impurity of Y2. is calculated as 0.05. Then, the improvement amount calculation unit 102d adds the impurity improvement amount (0) for Y1 and the impurity improvement amount (0.05) for Y2, so that the total impurity improvement amount is 0.05. 05.

(2-2) When dividing by X2 When dividing by X2, as shown in FIG. A record group (data of X2=1) consisting of four records (id=a, c, d, h) having X2=2 and six records having X2=2 (id=b, e, f, g, i , j) (X2=2 data).

Here, the impurity calculation unit 102f calculates the impurity of Y1 in the data of X2=1 as 0 as shown in FIG. 5 by the calculation method described in (1) above, and The impurity of Y2 is calculated as 0.625, the impurity of Y1 in the data of X2=2 is calculated as 0.4444, and the impurity of Y2 in the data of X2=2 is calculated as 0.7222.

Then, the impurity calculation unit 102f calculates a weighted average of the impurities calculated in the previous paragraph, for example, using the following calculation formula weighted by the number of cases. That is, weighted average={(impurity of data of X2=1)×(number of records in data of X2=1)/(number of records in data before division)}+{(impurity of data of X2=2) purity)×(the number of records in the data of X2=2)/(the number of records in the data before division)}. According to this formula, the impurity calculation unit 102f calculates the impurity after division for Y1 as (0×4/10)+(0.4444×6/10)≈0.2667. Similarly, the impurity calculation unit 102f calculates the impurity after division for Y2 as (0.625×4/10)+(0.7222×6/10)≈0.6833.

Finally, the improvement amount calculation unit 102d subtracts the impurity after the division of Y1 (0.2667) from the impurity before the division of Y1 (0.48) to obtain the improvement amount of the impurity of Y1. is calculated as 0.2133. Similarly, the improvement amount calculation unit 102d subtracts the impurity after the division of Y2 (0.6833) from the impurity before the division of Y2 (0.53) to obtain the improvement amount of the impurity of Y2. is calculated as -0.153. Then, the improvement amount calculation unit 102d adds the impurity improvement amount (0.2133) for Y1 and the impurity improvement amount (−0.153) for Y2, so that the total impurity improvement amount is is calculated as 0.06.

(2-3) When dividing by X3 When dividing by X3, the record dividing unit 102a divides 10 records in the analysis target data 106a of FIG. From a record group (data of X3=1) consisting of two records (id=a, b) having X3=2 and six records (id=c, d, e, f, g, h) having X3=2 and a record group (data of X3=3) consisting of two records (id=i, j) having X3=3.

Here, the impurity calculation unit 102f calculates the impurity of Y1 in the data of X3=1 as 0.5 as shown in FIG. The impurity of Y2 in the data is calculated as 0, the impurity of Y1 in the data of X3 = 2 is calculated as 0.5, the impurity of Y2 in the data of X3 = 2 is calculated as 0.5, and X3 = The impurity of Y1 in the data of 3 is calculated as 0, and the impurity of Y2 in the data of X3=3 is calculated as 0.

Then, the impurity calculation unit 102f calculates a weighted average of the impurities calculated in the previous paragraph using the calculation formula described in (2-2). That is, the impurity calculation unit 102f calculates the impurity after division for Y1 as (0.5×2/10)+(0.5×6/10)+(0×2/10)=0.4 to calculate. Similarly, the impurity calculation unit 102f calculates the impurity after division for Y2 as (0×2/10)+(0.5×6/10)+(0×2/10)=0.3. calculate.

Finally, the improvement amount calculation unit 102d subtracts the impurity after the division of Y1 (0.4) from the impurity before the division of Y1 (0.48) to obtain the improvement amount of the impurity of Y1. is calculated as 0.08. Similarly, the improvement amount calculation unit 102d subtracts the impurity after the division of Y2 (0.3) from the impurity before the division of Y2 (0.53) to obtain the improvement amount of the impurity of Y2. is calculated as 0.23. Then, the improvement amount calculation unit 102d calculates the total impurity improvement amount by adding the impurity improvement amount (0.08) for Y1 and the impurity improvement amount (0.23) for Y2. It is calculated as 0.31.

(2-4) Summary As calculated in (2-1) to (2-3) above, the sum of the Y1 impurity improvement amount and the Y2 impurity improvement amount is divided based on X1. was 0.05, 0.06 when divided based on X2, and 0.31 when divided based on X3. It was found that the gain is large).

In this example, focusing on the "total" of the impurity improvement amount of the objective variables (Y1 and Y2), X3 was adopted among the explanatory variables (X1, X2 and X3). Focusing on the "maximum value" of the improvement amount of the variables (Y1 and Y2), it may be determined which of the explanatory variables (X1, X2 and X3) to adopt. In this example, the impurity improvement amount for Y1 when divided by X1 is 0, the impurity improvement amount for Y2 when divided by X1 is 0.05, and the impurity for Y1 when divided by X2 Improvement amount of 0.2133, improvement amount of impurity for Y2 when divided by X2 -0.153, improvement amount of impurity for Y1 when divided by X3 0.08 and when divided by X3 Of the six improvement amounts of the impurity improvement amount of 0.23 for Y2, the impurity improvement amount of 0.23 for Y2 when divided by X3 corresponds to the "maximum value", so X3 is adopted. The Rukoto. In this way, by focusing on the "maximum value", the condition (explanatory variable) makes it possible to perform a good split (split that improves impurity) only for a specific objective item sequence (specific objective variable). Even if it is a thing, it can be made easier to be adopted.

(3) Divide the data by the value of the condition item with the largest information gain Next, the record division unit 102a is the condition item (objective variable) with the largest information gain described in (2-4). Using X3 as a reference, the records in the analysis target data 106a of FIG. 2 are divided for each value of X3.

Specifically, as shown in FIG. 7, the record dividing unit 102a divides 10 records in the analysis target data 106a of FIG. 2 from two records (id=a, b) having X3=1. A record group A consisting of 6 records (id=c, d, e, f, g, h) having X3=2, and a record group D consisting of 2 records (id=i , and j).

(4) Further repeat division for each of the divided data The record dividing unit 102a further repeats division for each of the three record groups (record groups A, D and I) obtained by dividing in (3). .

Specifically, as shown in FIG. 8, the record dividing unit 102a divides two records in the record group A into a record group B consisting of one record (id=a) having X2=1, =2 and a record group C consisting of one record (id=b). As a result of the division up to this point, each record group contains only one record, that is, there are no more records that can be divided, so the record dividing unit 102a stops dividing.

Also, as shown in FIG. 8, the record dividing unit 102a divides the six records in the record group D into a record group E consisting of three records (id=c, d, e) having X1=1, and a record group F consisting of three records (id=f, g, h) with X1=2. Furthermore, as shown in FIG. 8, the record dividing unit 102a divides the three records in the record group E into a record group G consisting of two records (id=c, d) having X2=1, and a record group G consisting of X2=1. 2, and a record group H consisting of one record (id=e) having 2.

Note that two records in the record group I have both X1 equal to 2 and X2 both equal to 2, and since there is no target variable that serves as a criterion for division, the division is terminated.

(5) Stop division based on some termination conditions As described in (4), the record division unit 102a performs recursive division processing. to stop.

One of the termination conditions is the number of data (the number of records in the group of records after division). Specifically, when the number of records in the record group after division is small (for example, when 5% of the analysis target data 106a is set as a threshold and the number of records in the record group after division is equal to or less than the threshold), It is conceivable to take measures such as stopping further division.

The second termination condition is the depth (the number of divisions performed). Specifically, if the number of divisions executed exceeds a preset number, further divisions can be stopped, thereby preventing the conditions from becoming too complicated.

A third termination condition is the improvement of impurities. Specifically, if the improvement in impurity is small as a result of division, further division may be stopped.

(6) Calculating the amount of improvement for each division When dividing the data, it is desirable to divide the data so that the objective variables can be more clearly distinguished. Impurities are not improved. For this reason, as will be described below, by calculating the impurity improvement amount for each objective variable, it is possible to grasp how much the division affects the objective variable.

For example, as shown in FIG. 9, when division 1 is performed based on X3=1, and division 2 is performed based on X2=1, the improvement amount calculation unit 102d performs division 1 as follows. and division 2, the improvement amount is calculated. Note that FIG. 9 is obtained by cutting out the record group and B in FIG.

Consider division 1 (division under the condition that X3=1). The impurity of Y1 before division 1 is 0.48, and the impurity of Y1 after division 1 is 0.5. is calculated as 0.48-0.5=-0.02. Similarly, the impurity of Y2 before division 1 is 0.53, and the impurity of Y2 after division 1 is 0. The purity improvement is calculated as 0.53-0=0.53. From this result, it can be seen that division under the condition of X3=1 has no effect on Y1, but has an effect on Y2. The results are shown in FIG.

Consider partition 2 (a partition under the condition that X3=1 and X2=1). The impurity of Y1 before division 2 is 0.5, and the impurity of Y1 after division 2 is 0. Therefore, the improvement amount calculation unit 102d improves the impurity of Y1 before and after division 2. The amount is calculated as 0.5-0=0.5. Similarly, the impurity of Y2 before division 2 is 0, and the impurity of Y2 after division 2 is 0. Therefore, the improvement amount calculation unit 102d calculates the impurity of Y2 before and after division 2. The amount of improvement is calculated as 0-0=0. From this result, it can be seen that division under the condition of X3=1 and X2=1 is effective for Y1, but not for Y2. The results are shown in FIG.

The amount of impurity improvement calculated in this way is retained as the division influence score.

(7) Output of analysis results From the results up to (6), it is possible to obtain the conditions for the feature items (explanatory variables) and the rules for the values that the objective variables should take at that time.

(7-1) Output by condition (by explanatory variable) Specifically, the ratio calculation unit 102b sets the ratio of Y1=2 to 1/1×100=100% for the group of records indicated by C in FIG. Then, the ratio of Y2=1 is also calculated as 1/1×100=100%. Similarly, the ratio calculation unit 102b calculates the ratio of Y1=1 as 2/2×100=100% for the record group indicated by G in FIG. %. Furthermore, in the same way, the ratio calculation unit 102b calculates the ratio of Y1=1 as 2/2×100=100% and the ratio of Y2=4 as 2/2×100=100% for the record group indicated by I in FIG. Calculate as 100%. Then, the analysis result output unit 102e outputs the result of the calculation as shown in FIG. In the output shown in FIG. 11, the ratio indicates the ratio of the data used for learning, and the display is such that there is no overlapping data for each row.

Furthermore, by using the division influence score calculated in (6), it is possible to select to output only highly effective conditions. For example, consider partitions 1 and 2 described in (6). Suppose that the condition is that "impurity improvement amount for objective variable is 0.05 or more". As explained in division 1 of (6), in the division under the condition of X3=1, the impurity improvement amount for Y1 is -0.02, and the impurity improvement amount for Y2 is 0.53. Also, as explained in division 2 of (6), in the division under the condition of X3=1 and X2=1, the impurity improvement amount for Y1 is 0.5, and the impurity for Y2 is The amount of improvement in purity was zero. Therefore, there are two combinations of explanatory variables and objective variables that meet the above conditions: a combination of "X3=1 and Y2" and a combination of "X3=1, X2=1 and Y1".

The analysis result output unit 102e displays only the two combinations described in the previous paragraph. That is, as shown in FIG. 12, the analysis result output unit 102e has a row indicating that the ratio of Y2=1 when X3=1 is 100% and a row indicating that Y1=1 when X3=1 and X2=1. and a line indicating that the percentage is 100%. This allows the user to check only objective variables that have been meaningfully divided in a list. In addition, in the display of FIG. 12, there is a possibility that there is data that corresponds to multiple lines of conditions (multiple explanatory variables), but compared to the case where the results in (6) are not used, simple conditions A combination of (explanatory variables) and objective variables can be presented to the user.

(7-2) Output by tree The results up to (6) can also be expressed in the form of a tree (decision tree) as follows.

Specifically, the ratio calculation unit 102b calculates the ratio of Y1=1 to be 6/10×100=60%, and the ratio of Y2=2 to 3/10×100= Calculate 30%. Similarly, for the record group indicated by A in FIG. %. Furthermore, in a similar manner, the ratio calculation unit 102b calculates ratios as shown in FIG. 13 for the record groups indicated by B to I in FIG. Then, the analysis result output unit 102e displays the result of the calculation as shown in FIG. In addition, in the table of FIG. 13, as the value of the objective variable, the mode value (for each value of the explanatory variable) under the condition and the ratio of the mode value are described.

Furthermore, the analysis result output unit 102e displays results that are easy for the user to understand and use by displaying only those with a large impurity improvement amount for the target variable or those with an impurity after division that is less than a certain value. can do FIG. 14 shows an image example of the tree of the result display. If the result is displayed, the user only needs to pay attention to the part where the value of the objective variable is input, and it is possible to grasp where the value of the objective variable is determined with a small number of conditions (explanatory variables). can be done. In addition, by showing the mode of the objective variable and the proportion of that value for each condition (explanatory variable value) in a tree, the user can see that the tree shown in the result display certainly matches the division result in FIG. You can recognize that they match.

(8) Example of application of algorithm to payroll system The decision tree algorithm described in this item [3-1] can It can be applied to registration and correction processes.

As the explanatory variables X1, X2, and X3 explained in this item [3-1], items such as employee classification (manager, general employee, part-time job, etc.), gender, position, work location, and qualifications are used. be able to. This item is an item for specifying an employee's status, is determined outside the system, and is an item whose value to be input to the system is clear.

As the objective variables Y1 and Y2 described in item [3-1], for example, items such as salary category, salary payment date category, working condition category, insurance category, and existence of various allowances can be used. The item is an item for determining the method of calculating or paying the salary to the employee, and is used in the salary system for calculating the salary. This item is determined and input based on the rules of

By displaying the analysis results up to (6) by the display method described in (7), it is possible to suggest what to input as the objective variable. Thereby, for example, erroneous salary payment can be prevented.

[3-2. Employee information registration process]
Next, employee information registration processing will be described in detail with reference to FIGS. 21 to 28. FIG.

In this section, (1) first, the processing up to the registration of the contents of the definition master 106b will be explained based on FIG. 21 with reference to FIGS. 23 to 26; The employee information registration processing performed after the registration of .

(1) Processing up to registering the contents of the definition master 106b (1-1) About "Analysis Execution" in FIG. 21 First, "Analysis Execution" in FIG. 21 will be described in detail. As shown in FIG. 21, when the analysis execution screen is displayed and "analysis is executed", the existing large amount of data in the employee data 106a1 is analyzed by the analysis system, and the analysis result is "candidates for simultaneous change displayed on the screen." item is displayed. An example of the analysis execution screen is shown in FIG. When "execute" is selected after specifying the target year and month on the analysis execution screen, the information in the employee data 106a1 that matches the specified target year and month becomes the target of analysis. Also, the analysis system is not particularly limited, and may be, for example, the analysis method described in [3-1] or other known analysis methods.

(1-2) “Display of Candidate Items for Simultaneous Change on Screen” in FIG. 21 Next, “Display of Candidate Items for Simultaneous Change on Screen” in FIG. 21 will be described in detail. The registration candidate pattern display unit 102g displays the result of the analysis on the pattern registration screen. Specifically, in the pattern registration screen shown in FIG. The four branches of "sales registrant" are checked and selected. "Aggregation of all conditions" is an item that displays the aggregation results of all cases to be analyzed, and in the case of analysis of employee information and registered definitions, it is set to be checked by default.

The registration candidate pattern display unit 102g identifiably displays items with a matching rate of 90% or more for the content selected in the conditional branch shown in the left part of FIG. 24 as registration targets. That is, the registration candidate pattern display unit 102g displays the salary item value having the largest ratio in the same salary item for the group of employees specified by the selected branch, together with the largest ratio, as shown in FIG. Displayed on the pattern registration screen. At this time, the registration candidate pattern display unit 102g displays the salary item value with the largest ratio by hatching with oblique lines when the largest ratio is 100%, and by hatching with dots when the largest ratio is 90% or more. Then, the pattern registration screen of FIG. 24 is displayed. In other words, the registration candidate pattern display unit 102g performs display with a threshold of 90%, but the threshold can be arbitrarily changed by the operator's setting.

Specifically, the registration candidate pattern display unit 102g displays, as indicated by MA4 in FIG.
● For the branch of "All conditions tabulation", "Salary category = 4, 83%", "Salary payment date category = 2, 83%", "Bonus payment date category = 1, 99%", "Work conditions Category = 2, 83%", "Health insurance category = 0, 70%", "Employees' pension category = 0, 70%", "Social insurance collection category = 0, 100%", "Social insurance acquisition category = 1, 81%"","Social insurance loss category = 0, 100%", "Insurance type = 0, 62%", "Employment insurance category = 0, 68%", "Worker's accident insurance category = 1, 70%", and "Paid grant table Classification = 1, 52%”, and among these, “social insurance collection classification = 0, 100%” and “social insurance loss classification = 0, 100%” with a ratio of 100% are hatched as candidates for registration. , and "bonus payment date division = 1, 99%" with a ratio of 90% or more is displayed with dot hatching as a candidate for registration ("social insurance acquisition division = 1, 81%" It is not attached in this process, but is attached by the process from "adjustment by item" described later),
● For the “employee category = general employee” branch, “salary category = 2, 100%”, “salary payment date category = 1, 100%”, “working condition category = 1, 100%”, “health Insurance classification = 1, 99%", "Employees' pension classification = 1, 99%", "Insurance classification = 1, 86%", "Employment insurance classification = 1, 100%", "Workers' accident insurance classification = 1, 100%" ” and “Paid grant table classification = 1, 95%”, among these, “Salary classification = 2, 100%”, “Salary payment date classification = 1, 100%”, "Labor conditions category = 1, 100%", "Employment insurance category = 1, 100%", and "Workers' accident insurance category = 1, 100%" are indicated by diagonal hatching as candidates for registration. "Health insurance classification = 1, 99%", "Employees' pension classification = 1, 99%", and "Paid grant table classification = 1, 95%" are indicated by hatching dots as candidates for registration,
●For the branch of "gender = male", "insurance type = 1, 99%" with a ratio of 90% or more is displayed as a candidate for registration by hatching dots.
● For the branch of "business = sales registrant", "paid grant table division = 1, 98%" with a ratio of 90% or more is displayed by hatching dots as registration candidates.
To briefly explain what this indication means, for example, in the case of a group of employees who fall under the category of "employee classification = general employee" and "gender = male", the percentage of employees whose insurance type is 1 is 99%. That is.

The registration candidate pattern display unit 102g may display the salary item values hatched as registration candidates in MA4 of FIG. 24 as they are in MA3 of FIG. Only the salary item values selected by the operator from the "adjustment by item" button indicated by MA2 on the upper right of the screen may be displayed in MA3 of FIG.

(1-3) “Select from Candidates” in FIG. 21 Finally, “Select from Candidates” in FIG. 21 will be described in detail. When "adjustment by item" indicated by MA2 on the upper right of the screen in FIG. 24 is selected, the registration candidate pattern display section 102g displays the salary item values displayed in MA4 in FIG. 24 as indicated by MA5 in FIG. Show a checkbox. Here, the registration candidate pattern display unit 102g checks the check boxes corresponding to the salary item values hatched in MA4 of FIG. It may be displayed as is. Specifically, in MA5 of FIG. 24, the registration candidate pattern display unit 102g selects the check boxes corresponding to "social insurance collection category=0, 100%" and "social insurance loss category=0, 100%". It is displayed with a check from the beginning. In addition, in MA5 of FIG. 25, check boxes corresponding to all "salary item=salary item value, ratio" explained as being hatched in the paragraph two before are actually displayed. is cut and displayed.

Then, the operator can uncheck the checked checkboxes and, conversely, check the unchecked checkboxes. This means that the operator can select any salary item value. In MA5 of FIG. 24, the check box corresponding to "social insurance acquisition category=1, 81%" was initially unchecked, but was selected and checked by the operator. Note that the registration candidate pattern display unit 102g may display the salary item value corresponding to the selected check box with horizontal hatching, as indicated by MA5 and MA4 in FIG. In other words, the registration candidate pattern display unit 102g may display items whose matching rate exceeds the threshold value (90% in this example) and items selected from the "item-by-item adjustment" by hatching. .

Finally, the registration candidate pattern display unit 102g displays the salary item values corresponding to the checked check boxes (including both checked boxes as initial values and checked boxes by the operator), As a final registration candidate, it is displayed in the "selection result" portion indicated by MA3 in FIG. Specifically, as indicated by MA3 in FIG. 24, the registration candidate pattern display unit 102g displays "salary classification=2", "salary payment date classification=1", and "bonus payment date" as final registration candidates. Category = 1”, “Working condition category = 1”, “Health insurance category = 1”, “Employees’ pension category = 1”, “Social insurance collection category = 0”, “Social insurance acquisition category = 1”, “Social insurance loss category = 0", "insurance type=1", "employment insurance category=1", "worker's accident insurance category=1" and "paid grant table category=1" are displayed.

Then, when the definition name "general male employee (sales registrant)" is set for the pattern in MB1 of FIG. 25, and the "registration" button in MB2 of FIG. , one pattern is registered in the definition master 106b. That is, the pattern registration unit 102h registers the status items and the values of the status items for the branch selected in the condition selection, namely, "employee classification=general employee", "sex=male", "duty=no input", and "duty=sales". registrant", and salary items and salary item values displayed as candidates for final registration by the registration candidate pattern display unit 102g, ie, "salary classification=2", "salary payment date classification=1", and "bonus payment date Category = 1”, “Working condition category = 1”, “Health insurance category = 1”, “Employees’ pension category = 1”, “Social insurance collection category = 0”, “Social insurance acquisition category = 1”, “Social insurance loss category = 0", "Insurance type = 1", "Employment insurance classification = 1", "Industrial accident insurance classification = 1", and "Paid grant table classification = 1" (hereinafter referred to as "Pattern 1"). is defined as "definition number: 100, definition name: male general employee (sales registrant)" and registered in the definition master 106b.

Note that the registered pattern can be confirmed and edited on the screen of FIG. 26 (having the same configuration as the screen for registering employee information). Specifically, at the location indicated by MC1 in FIG. 26, the values to be updated at the same time can be edited.

As described in (1) above, according to the information processing apparatus 100 according to the present embodiment, patterns of salary item values that are candidates for registration are shown by hatching with oblique lines and dots in MA4 in FIG. can be displayed together with the ratio, the operator can easily and accurately register patterns of salary item values while referring to the display. Further, the displayed pattern may be displayed as it is in MA3 of FIG. 24, but in the information processing apparatus 100 according to the present embodiment, it is possible to select the "adjustment by item" button indicated by MA2 in FIG. After performing detailed manual adjustments (=unchecking or checking) for each item (salary item value unit), the patterns of salary item values that are candidates for final registration are displayed in MA3 of Fig. 24. Therefore, the pattern of salary item values can be flexibly adjusted according to the operator's intention and then registered.

(2) Employee Information Registration Process Performed After Registering the Contents of the Definition Master 106b First, as shown in FIG. 22, a preset employee registration pattern is selected. Specifically, when the MD1 button is selected as shown in the employee information registration/change screen in FIG. 27, another screen in FIG. Multiple definitions are called and displayed. In this example, the definition "definition number: 100, definition name: male general employee (sales registrant)" is designated as indicated by MD2 in FIG. 27, and the "execute" button indicated by MD3 in FIG. 27 is pressed. shall be In the employee information registration/change screen of FIG. 27, the definition may be specified by directly inputting the definition number in the blank shown to the immediate left of MD1. In other words, the definition may be specified by selecting from multiple definitions displayed on a separate screen from the employee information registration/change screen, or by direct input on the employee information registration/change screen. may

Subsequently, as shown in FIG. 22, the contents of the pattern are displayed on the employee information registration/change screen as initial values. Specifically, from the definition master 106b, the registration candidate item display unit 102i selects the pattern 1 (Consists of an identity item and an identity item value and a salary item and a salary item value. See (1-3) for details.). Then, as shown in ME2 in FIG. 28, the registration candidate item display unit 102i displays the acquired status item value "2 (regular employee)" for the acquired status item "employee classification" and the acquired salary item "salary "2" for the acquired salary item value "2" for the "Bonus Payment Date Category", the acquired salary item value "1" for the acquired salary item "Salary Payment Date Category", and the acquired salary item "Bonus Payment Date Category" The acquired salary item value "1" is displayed on the employee information registration/change screen as a candidate for new registration. In other words, in ME2 of FIG. 28, items are automatically set according to the contents of the pattern. In addition, since the portion indicated by ME1 in FIG. 28 is a portion that requires individual input, different information is input for each employee.

Finally, as shown in FIG. 22, the contents displayed on the employee information registration/change screen are registered in the employee data 106a1 as employee information. Specifically, when the "registration" button indicated by ME3 in FIG. 28 is pressed, the employee information displayed on the employee information registration/change screen in FIG. Date of birth, date of employment, employee classification: 2 (regular employee), salary classification: 2, salary payment date classification: 1, and bonus payment date classification: 1" are registered in the employee data 106a1.

As described in (2) above, according to the information processing apparatus 100 according to the present embodiment, the operator simply designates a definition on the employee information registration/change screen, and as shown in ME2 in FIG. Candidates for registration of information (identification item values and salary item values) can be displayed, so the operator does not need to input employee information from scratch, and as a result, it is possible to easily and accurately register employee information. Become.

[3-3. Change processing of employee information]
Finally, employee information change processing will be described in detail with reference to FIGS. 21 and 33 to 39. FIG.

In this section, (1) first, the processing up to the registration of the contents of the definition master 106b will be described with reference to FIGS. 34 to 36 based on FIG. 21; will be described with reference to FIGS. 37 to 39 based on FIG. 33. FIG.

(1) Processing until the contents of the definition master 106b are registered The processing contents until the contents of the definition master 106b are registered are the same as (1) in [3-2], so a detailed explanation is omitted. Only different points will be briefly explained.

In the pattern registration screen shown in FIG. 34, among the conditional branching branches shown on the left side, there are four "employee classification = part-time job", "duty = no input", "gender = female", and "task = sales registrant". One branch is checked and selected.

The registration candidate pattern display unit 102g, as shown in MF4 in FIG.
● For the branch of "employee classification = part-time job", "salary classification = 4, 100%", "salary payment date classification = 2, 100%", "working condition classification = 4, 100%", "health insurance Category = 0, 80%", "Employees' Pension Category = 0, 80%", "Insurance Category = 0, 68%", "Employment Insurance Category = 1, 78%", and "Workers' Accident Insurance Category = 1, 80%" , and among these, "salary category = 4, 100%", "salary payment date category = 2, 100%" and "working conditions category = 4, 100%" with a ratio of 100% are candidates for registration. is indicated by diagonal hatching as
● For the branch of "Gender = Female", "Insurance type = 2, 53%" is displayed,
● For the branch of "business = sales registrant", "insurance type = 2, 94%", "employment insurance category = 1, 100%", "workmen's accident insurance category = 1, 100%" and "paid Among these, "employment insurance category = 1, 100%" and "workmen's accident insurance category = 1, 100%", which have a ratio of 100%, are indicated by diagonal lines as candidates for registration. , and "insurance type=2, 94%" with a ratio of 90% or more is displayed as a candidate for registration by hatching dots.

Then, the registration candidate pattern display unit 102g may display the salary item values hatched as registration candidates in MF4 of FIG. 34 as they are in MF3 of FIG. Only the salary item values selected by the operator from the "adjustment by item" button indicated by MF2 on the upper right of the screen may be displayed in MF3 of FIG.

Finally, the definition name "female part-time employee (sales registrant)" is set as shown in MG1 in FIG. 35 for the pattern displayed in MF3 in FIG. When the "registration" button in MG2 of 35 is selected, the pattern registration unit 102h registers one pattern in the definition master 106b. That is, the pattern registration unit 102h registers the status item and the value of the status item for the branch selected in the condition selection, namely, "employee classification = 10 (part-time job)", "duty = no input", "sex = female" and "work = sales registrant", and the salary items and salary item values displayed as final candidates for registration by the registration candidate pattern display unit 102g, "salary classification=4", "salary payment date classification=2", "labor A pattern (hereinafter referred to as “pattern 2”) consisting of “condition classification = 4”, “insurance classification = 2”, “employment insurance classification = 1” and “workmen’s accident insurance classification = 1” is defined as “definition number: 700 , definition name: female part-time employee (sales registrant)” and registered in the definition master 106b.

Note that the registered pattern can be checked and edited on the definition information registration/change screen of FIG. Specifically, at the location indicated by MH1 in FIG. 36, the values to be updated at the same time can be edited.

(2) Employee Information Change Processing After Registering Contents of Definition Master 106b First, as shown in FIG. 33, employee information is displayed by referring to the employee data 106a1. Specifically, when "1000000001" is specified as the employee code on the employee information registration/change screen of FIG. to be displayed.

Subsequently, as shown in FIG. 33, the employee classification is manually changed. Specifically, as indicated by MI1 in FIG. 37, the employee classification is changed from "4 (regular employee)" to "10 (part-time employee)".

When the change is made, as shown in FIG. 33, the definition master 106b is referred to confirm whether there is a simultaneous change candidate item, and a simultaneous change pattern is displayed on the screen. Specifically, when the employee classification “4” explained in the previous paragraph is changed to “10”, the pattern definition display unit 102j displays the changed target status item “employee classification” from the definition master 106b. ” and the changed status item value “10”, acquire “female part-time employee (sales registrant)” and “temporary employee (unqualified person)”. Then, as shown in the upper screen of FIG. 38, the pattern definition display unit 102j displays the acquired definition name "female part-time employee (sales registrant)" as an identification item and an identification item value (employee classification: 10, job title: -1, gender: 2, business: 70), and the acquired definition name "Temporary employee (non-qualified person), the status item and status item value associated with it ( Employee category: 10, Salary category: 1, Place of work: -1, Qualification: NOT 70, Occupation: 10) In this way, the pattern definition display unit 102j displays the "Condition item ” column displays definitions that include changed items, but it is also possible to change the order of display according to the precision of the analysis results.

When "Confirm" indicated by MJ2 is selected on the upper screen of FIG. 38, the employee information can be confirmed on the definition information registration/change screen (screen when the definition is registered) shown in the lower part of FIG.

Finally, returning to FIG. 33, by selecting a specific pattern, it is possible to automatically change a plurality of items corresponding to the selected pattern at the same time and update the contents of the employee data 106a1. Specifically, when the definition name “female part-time employee (sales registrant)” is designated by selecting “execute” indicated by MJ3 on the upper screen of FIG. From the definition master 106b, as a pattern associated with the specified definition name "female part-time employee (sales registrant)", pattern 2 (composed of an identity item, an identity item value, a salary item, and a salary item value. Details See (1).) to obtain the item and the value. 39, the change candidate item display unit 102k displays the acquired salary item value "4" for the acquired salary item "salary category" and the acquired salary item "salary payment date category". ” is displayed on the employee information registration/change screen as a change candidate. In other words, the change candidate item display unit 102k not only automatically changes the definition set item (=indicated by MK2 in FIG. 39), but also the remaining items (indicated by MK3 in FIG. 39) other than the condition item changed this time. ) should also need to be changed, so it can be complemented and changed automatically. By registering the changed employee information in the employee data 106a1, the employee information of the employee with the employee code "1000000001" can be updated from that for regular employees to that for part-time workers.

As described in (2) above, according to the information processing apparatus 100 according to the present embodiment, the operator changes the employee classification from "4 (regular employee)" to "10 (part-time employee)" on the employee information registration/change screen. Only by changing, as shown in MK1 in FIG. 39, the salary division is automatically changed from "1" to "4", and the salary payment date division is automatically changed from "1" to "2". could be displayed. That is, according to the information processing apparatus 100 according to the present embodiment, when changing employee information, the operator only needs to change one item and change other items (usually there are a huge number of items). No need to change. As a result, it is possible to easily and accurately change employee information.

(3) Other Examples of Utilization of Changing Employee Information In the above, (1) and (2) were explained using the example of "change of status" as a scene for changing employee information. can be used, for example, in the following scenes.

The first example is "change of affiliation". This is because it is assumed that there are items that should be changed at the same time because the handling of employees may change due to a change in the department to which they belong.

A second example is a change in work location. This is because there are special allowances (cold area allowances) depending on the place of work, so it is assumed that there are items that should be changed at the same time.

A third example is “childcare leave, maternity leave, etc.” It is assumed that there are items that should be changed at the same time, because there are special rules unique to the company, such as no allowance being made or special allowance being made. In addition, it is necessary to change a plurality of items managed by the system according to the social insurance license.

[4. Summary of this embodiment]
[4-1. Decision tree summary]
Thus, according to the information processing apparatus 100 according to the present embodiment, as described in [1-1] and [3-1], even if the data to be analyzed includes a plurality of objective variables, 1 Data can be analyzed in a form that can be collectively output as one analysis result.

Here, the decision tree algorithm is known as an algorithm for analyzing the rule of the item, but the existing decision tree algorithm cannot handle a plurality of objective variables. was improved to handle multiple objective variables. In addition, in the present embodiment, for example, by scoring which item (which explanatory variable) value influences the determination of which item (which objective variable) value, the meaning can be interpreted by the user. Easy display is possible. The decision tree algorithm according to this embodiment can be used, for example, in the following situations.

For example, when registering business data, if one item is changed, other items may need to be changed as well. In this embodiment, the rules between items can be derived from existing data. Specifically, by improving the existing decision tree algorithm, simple presentation of rules between items (between the certain item and the other item) has been realized. In addition, for example, if the operation rules vary or are not clear for each individual company, it may be difficult to pre-configure the system. In this embodiment, by deriving rules using existing data, it can be used for support and checking at the time of subsequent input, and as a result, it leads to reduction of input errors and improvement of input efficiency.

[4-2. Summary of registering and changing employee information]
Thus, according to the information processing apparatus 100 according to the present embodiment, as described in [1-2], [3-2], and [3-3], the status item value (for example, the employee classification is regular employee or part-time job, gender is male or female, etc.), and the salary item value to be registered for the selected status item value (for example, what is the salary category, what is the insurance type? , What is the employment insurance classification, etc.) can be displayed on the screen to support the operation of registering the patterns.

Here, there are a huge number of items in employee information that should be set (that is, newly registered) when calculating salaries. Also, when changing one item, there are often many items to be changed at the same time due to salary regulations. However, since salary regulations often differ from company to company, it was not possible to newly register or change items on the system in the past. For this reason, it is necessary to manually register new items and change items.

Therefore, in the present embodiment, for example, by AI analysis of a large amount of employee data registered in the system, it is possible to propose items to be registered and items to be changed when registering and changing values of employee information. made it In this way, the system proposes the content of registration and changes in accordance with the company's salary regulations, thereby preventing, for example, omissions in registration, mistakes in registration, omissions in changes, and mistakes in changes, etc., and erroneous payment of salaries. be able to. Furthermore, for example, registration work and change work can be facilitated.

[5. Other embodiments]
The present invention may be implemented in various different embodiments other than the embodiments described above within the scope of the technical idea described in the claims.

For example, among the processes described in the embodiments, all or part of the processes described as being automatically performed can be manually performed, or all of the processes described as being manually performed Alternatively, some can be done automatically by known methods.

In addition, unless otherwise specified, the processing procedures, control procedures, specific names, information including parameters such as registration data and search conditions for each process, screen examples, and database configurations shown in this specification and drawings can be changed arbitrarily.

Also, with respect to the information processing apparatus 100, each illustrated component is functionally conceptual, and does not necessarily need to be physically configured as illustrated.

For example, the processing functions of the information processing apparatus 100, particularly the processing functions performed by the control unit, may be implemented in whole or in part by a CPU and a program interpreted and executed by the CPU. Alternatively, it may be implemented as hardware by wired logic. The program is recorded on a non-temporary computer-readable recording medium containing programmed instructions for causing the information processing apparatus to execute the processing described in this embodiment, and the information processing apparatus as necessary. 100 mechanically read. That is, a storage unit such as a ROM or HDD (Hard Disk Drive) stores a computer program for giving commands to the CPU in cooperation with the OS to perform various processes. This computer program is executed by being loaded into the RAM and constitutes a control section in cooperation with the CPU.

Further, this computer program may be stored in an application program server connected to the information processing apparatus 100 via any network, and all or part of it can be downloaded as necessary. be.

Also, the program for executing the processing described in this embodiment may be stored in a non-temporary computer-readable recording medium, or may be configured as a program product. Here, the term "recording medium" refers to memory cards, USB (Universal Serial Bus) memories, SD (Secure Digital) cards, flexible disks, magneto-optical disks, ROMs, EPROMs (Erasable Programmable Read Only Memory), EEPROMs (registered (trademark) (Electrically Erasable and Programmable Read Only Memory), CD-ROM (Compact Disk Read Only Memory), MO (Magneto-Optical disk), DVD (Digital Versatile Disk), and Disc (registered trademark) such as Blu- shall include any "portable physical medium".

A "program" is a data processing method written in any language or writing method, regardless of the format such as source code or binary code. In addition, the "program" is not necessarily limited to a single configuration, but is distributed as multiple modules or libraries, or cooperates with a separate program represented by the OS to achieve its function. Including things. It should be noted that well-known configurations and procedures can be used for the specific configuration and reading procedure for reading the recording medium in each device shown in the embodiments, the installation procedure after reading, and the like.

The various databases stored in the storage unit are storage means such as memory devices such as RAM and ROM, fixed disk devices such as hard disks, flexible disks, and optical disks. It stores programs, tables, databases, files for web pages, and so on.

Further, the information processing apparatus 100 may be configured as an information processing apparatus such as a known personal computer or workstation, or may be configured as the information processing apparatus to which arbitrary peripheral devices are connected. Further, the information processing apparatus 100 may be implemented by installing software (including programs, data, etc.) that causes the apparatus to implement the processes described in the present embodiment.

Furthermore, the specific forms of distribution and integration of devices are not limited to those shown in the figures, and all or part of them can be functionally or physically arranged in arbitrary units according to various additions or functional loads. It can be distributed and integrated. In other words, the embodiments described above may be arbitrarily combined and implemented, or the embodiments may be implemented selectively.

INDUSTRIAL APPLICABILITY The present invention is useful in all industries and industries, but is particularly useful in the field of managing salary payments.

100 information processing device 102 control unit
102a record dividing unit
102b ratio calculator
102c Analysis result creation unit
102d improvement amount calculation unit
102e analysis result output unit
102f Impurity calculator
102g registration candidate pattern display section
102h Pattern registration unit
102i registration candidate item display area
102j Pattern definition display area
102k change candidate item display unit 104 communication interface unit 106 storage unit
106a Data to be analyzed
106a1 Employee Data
106b Automatic update definition master
106c Employee classification master
106d Salary category master
106e Job title master
106f grade master
106g Occupation master
106h Work location master
106i calculation definition master 108 input/output interface unit 112 input device 114 output device 200 server 300 network

Claims (8)

A data analysis device comprising a control unit,
The control unit
Record dividing means for dividing the records in the data to be analyzed including a plurality of records having an explanatory variable and a plurality of objective variables into a group of records based on the specific explanatory variable as a reference, and for each value of the explanatory variable used as the reference. When,
a ratio calculating means for calculating, for each of the record groups obtained by dividing by the record dividing means, the mode value in the same objective variable and the ratio of the mode value to all of the plurality of objective variables;
analysis result creation for creating an analysis result including the mode value and the ratio calculated by the ratio calculation unit for each of the values of the explanatory variable used as the basis for the division by the record division unit; means and
to provide
A data analysis device characterized by: The record dividing means
The division is repeated using the explanatory variable other than the explanatory variable used as the basis for the division as a new basis, and the number of the records in the record group obtained by the division, the number of times the division is performed, and the number before the division Stopping the division according to at least one of an impurity improvement amount that is an impurity improvement amount for the objective variable between and after the division;
The data analysis device according to claim 1, characterized by: The control unit
improvement amount calculation means for calculating a value obtained by subtracting the impurity after the division from the impurity before the division as the impurity improvement amount for each of the objective variables;
Analysis result output for outputting only the mode value and the ratio after the division for the objective variable for which the impurity improvement amount calculated by the improvement amount calculation means is equal to or greater than a predetermined value, among the analysis results. means and
further providing
3. The data analysis device according to claim 2 , characterized by: The control unit
Impurity calculation means for calculating the value obtained by subtracting the sum of the squares of the ratios of the values in the same objective variable from 1 as the impurity,
4. The data analysis device according to claim 2 or 3, characterized by: The analysis result creation means is
As the analysis result, the ratio calculating means for all of the plurality of objective variables for each branch hierarchically branched based on the explanatory variable used as the criterion for the division by the record dividing means and the value of the explanatory variable creating a decision tree including the calculated mode and the proportion;
5. The data analysis device according to any one of claims 1 to 4, characterized by: The explanatory variable is an item for identifying an employee's status,
The objective variable is an item for determining how to calculate or pay the salary to the employee;
6. The data analysis device according to any one of claims 1 to 5, characterized by: A data analysis method executed by an information processing device comprising a control unit,
executed by the control unit;
A record splitting step of splitting the records in the data to be analyzed including a plurality of records having an explanatory variable and a plurality of objective variables into a group of records based on the specific explanatory variable as a reference, for each value of the explanatory variable used as the reference. When,
a ratio calculating step of calculating, for each of the record groups obtained by dividing in the record dividing step, the mode value in the same objective variable and the ratio of the mode value to all of the plurality of objective variables;
creating an analysis result including the mode value and the ratio calculated in the ratio calculating step for each value of the explanatory variable used as the basis for the division in the record dividing step; a step;
including
A data analysis method characterized by: A data analysis program to be executed by an information processing device comprising a control unit,
for causing the control unit to execute
A record splitting step of splitting the records in the data to be analyzed including a plurality of records having an explanatory variable and a plurality of objective variables into a group of records based on the specific explanatory variable as a reference, for each value of the explanatory variable used as the reference. When,
a ratio calculating step of calculating, for each of the record groups obtained by dividing in the record dividing step, the mode value in the same objective variable and the ratio of the mode value to all of the plurality of objective variables;
creating an analysis result including the mode value and the ratio calculated in the ratio calculating step for each value of the explanatory variable used as the basis for the division in the record dividing step; a step;
including
A data analysis program characterized by:

Images