Decision tree discrimination support apparatus

Abstract

To support improvement of analysis accuracy by decision tree discrimination by securing readability of an analysis process.SOLUTION: A decision tree discrimination support apparatus 1 includes: a setting section 13 in which a decision tree is set from a threshold value of an adopted variable, and a prediction result of appropriateness or inappropriateness of a branch discriminated by the decision tree is set; an accuracy calculation section 14 that calculates analysis accuracy including precision and recall of the decision tree from appropriateness data of a corresponding data group and the prediction result; a recording section 15 that accumulates and records stratification conditions and analysis accuracy of a branch having the largest number of pieces of analysis data predicted to be appropriate among branches for each decision tree for each change of setting by the setting section 13; and an output section 16 that causes a display device to output the recorded stratification conditions and the analysis accuracy of the branch.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to a decision tree discrimination support device that supports analysis using decision tree discrimination. [Background technology]

[0002] As an example of this type of technology, Patent Document 1 proposes a multiple regression analysis device that outputs to a user information regarding the consistency between the multiple regression equation and the actual measured values ​​and information regarding the consistency between the integrated multiple regression equation and the actual measured values. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Publication No. 2022-064707 Summary of the Invention [Problem to be solved by the invention]

[0004] However, the purpose of the analysis device disclosed in Patent Document 1 is to design logic and a screen for considering improving the accuracy of linear multiple regression analysis by stratification. Therefore, in order to improve the analysis accuracy by decision tree discrimination, it is important for the user to understand changes in the thresholds of the adopted variables by the decision tree and changes in the analysis accuracy, and the readability of the analysis process cannot be guaranteed.

[0005] The present invention has been made in consideration of the above points, and an object of the present invention is to provide a decision tree discrimination support device that supports improvement of analytical accuracy through decision tree discrimination by ensuring readability of the analysis process. [Means for solving the problem]

[0006] In view of the above-mentioned problems, the present invention provides a decision tree discrimination support device that supports analysis by decision tree discrimination for all data groups having a plurality of analytical data each including attribute data relating to an attribute, values ​​of a plurality of adopted variables, and appropriate or inappropriate suitability data as a result of discrimination, the decision tree discrimination support device comprising: a storage unit that stores analytical data for the all data groups; an extraction unit that extracts, via an input device, from the stored all data groups, a corresponding data group corresponding to selected attribute data; and a decision tree that is set for the corresponding data group from the values ​​of the adopted variables and threshold values ​​of the adopted variables that serve as stratification conditions via the input device; The system is characterized by comprising: a setting unit that sets the prediction result of suitability or insuitability of the branch determined by the decision tree; an accuracy calculation unit that calculates the analysis accuracy consisting of the precision and recall of the decision tree from the suitability data of the corresponding data group and the prediction result set by the setting unit; a recording unit that, each time the setting is changed by the setting unit, accumulates and records the stratification conditions and the analysis accuracy of the branch having the largest number of analysis data predicted to be suitable among the branch of each decision tree; and an output unit that outputs the recorded stratification conditions and the analysis accuracy of the branch to a display device. [Effects of the Invention]

[0007] According to the present invention, each time the setting unit changes the settings, the display device can display the accumulated and recorded contents of the stratification conditions and analytical accuracy of the branch with the largest number of analytical data items predicted as appropriate among the branches of each decision tree. As a result, by ensuring the readability of the analysis process, it is possible to support the improvement of analytical accuracy by decision tree discrimination. [Brief explanation of the drawings]

[0008] [Figure 1] FIG. 1A is a block diagram of a decision tree discrimination support device, and FIG. 1B is a diagram showing an example of the entire data group of analysis data. [Figure 2] 2 is a diagram showing an example of a decision tree and analysis results displayed on a display device of the decision tree discrimination support device shown in FIG. 1. FIG. [Figure 3] 2 is a diagram showing the recorded contents of a recording unit displayed on a display device of the decision tree discrimination support device shown in FIG. 1. FIG. DETAILED DESCRIPTION OF THE INVENTION

[0009] A decision tree discrimination support device according to an embodiment of the present invention will be described below with reference to FIGS. 1 to 3. As shown in FIG. 1, the decision tree discrimination support device 1 according to this embodiment is a device that supports analysis, by decision tree discrimination, of a whole data group having a plurality of pieces of analytical data, each of which includes attribute data relating to one of a plurality of attributes, values ​​of a plurality of adopted variables linked to the attribute data, and suitability data indicating appropriateness or inappropriateness as a discrimination result. For example, the device supports a user in analyzing, by decision tree discrimination, evaluation results (such as tensile strength and surface hardness) of a manufactured material using adopted variables, such as the composition ratio of a material (e.g., metal material or resin material) and manufacturing conditions (e.g., heating time during manufacturing).

[0010] The system includes a processing device 10 that supports analysis of all data groups (described later) using decision tree discrimination. An input device 20 and a display device 30 are connected to the processing device 10. The processing device 10 includes, as hardware, a storage device (not shown) that stores all data groups and setting conditions, etc., and a processing device (not shown) that calculates analysis accuracy, etc. The processing device 10 includes, as software, a storage unit 11, an extraction unit 12, a setting unit 13, an accuracy calculation unit 14, a recording unit 15, and an output unit 16.

[0011] The storage unit 11 stores analysis data for all data groups. As shown in FIG. 1(b), the analysis data are Data 1 to Data N, and the entire data group includes all data from Data 1 to Data N. Each analysis data has attribute data for Attributes 1 and 2. In this embodiment, Attribute 1 includes A and B, and Attribute 2 includes A to C. Examples of attributes include the part name, use environment, and general name of the material that is the subject of the analysis data. In this embodiment, the attribute data associates values ​​X1 to XN, Y1 to YN, and Z1 to ZN with the adopted variables X, Y, and Z of Data 1 to Data N, which are the analysis data. For example, the adopted variables X, Y, and Z are variables such as the content of a specific composition of the material when manufacturing the material, the heating temperature, and the heating time. Furthermore, the analysis data, Data 1 to Data N, are associated with either "appropriate" or "inappropriate," which is the result of the judgment, as suitability data. "Appropriate" may be assigned a value of 1, and "inappropriate" may be assigned a value of 0. For example, if the tensile strength or surface hardness of the material in the analysis data is equal to or greater than a desired value, it is deemed to be appropriate, and if it does not satisfy the desired value, it is deemed to be inappropriate.

[0012] The extraction unit 12 extracts, from all stored data groups, a corresponding data group that corresponds to the attribute data selected via the input device 20. For example, as shown in Fig. 2, the user selects attribute data "A" for attribute 1 and attribute data "b" for attribute 2. In this case, 100 pieces of data are extracted as a corresponding data group from all data groups shown in Fig. 1(b).

[0013] The setting unit 13 sets a decision tree for the corresponding data group from the adopted variables and thresholds of the adopted variables that serve as stratification conditions via the input device 20, and sets a prediction result of the appropriateness or inappropriateness of the branch determined by the decision tree. For example, as shown in FIG. 2, a decision tree is set in which thresholds XA, XB, YA, YB, ZA, and ZB that are predicted to produce appropriate results are set for the adopted variables X, Y, and Z, and a prediction result for the analysis data at the end of the branch is set. In this example, there are four ends of the branch judged to be "appropriate," and the numbers of corresponding analysis data at each end are, from left to right, "4," "13," "4," and "1."

[0014] The accuracy calculation unit 14 calculates the analysis accuracy, which is made up of the precision and recall of the decision tree, from the suitability data of the corresponding data group and the set prediction results. Specifically, the number of analysis data considered to be appropriate in the stored suitability data is 40, and the number of analysis data considered to be inappropriate is 60. Meanwhile, of the prediction results predicted by the user using the decision tree, the number of analysis data considered to be appropriate is 22, and the number of analysis data considered to be inappropriate is 78. The accuracy calculation unit 14 creates a confusion matrix shown in the lower left of Figure 2 from the suitability data of the corresponding data group and the set prediction results, and calculates the precision and recall of the decision tree.

[0015] Each time the setting unit 13 changes the settings, the recording unit 15 cumulatively records the stratification conditions and analytical accuracy of the branch (main flow) of each decision tree that has the largest number of analytical data items predicted as appropriate. Specifically, this corresponds to the main flow of the branch of the decision tree shown in FIG. 2. For example, these results are cumulatively recorded as shown in FIG. 3, and the result shown in FIG. 2 is recorded as setting condition 1 in the fourth row of the table shown in FIG. 3. Other changed setting conditions and their results are also recorded in other rows. The output unit 16 outputs the recorded stratification conditions and analytical accuracy of the branch to the display device 30 in the format shown in FIGS. 2 and 3.

[0016] In this way, for each change in settings by the setting unit 13, the stratification conditions and analytical accuracy of the branch having the largest number of analytical data predicted as appropriate among the branches for each decision tree can be accumulated and recorded and displayed on the display device 30. As a result, the user can easily track the improvement in accuracy and changes in adopted variables before and after stratification in the discriminant analysis, and can make efficient analytical decisions.

[0017] Although the embodiments of the present invention have been described in detail above, the present invention is not limited to the above-described embodiments, and various design modifications can be made without departing from the spirit of the present invention as set forth in the claims. [Explanation of symbols]

[0018] 1: decision tree discrimination support device, 11: storage unit, 12: extraction unit, 13: setting unit, 14: accuracy calculation unit, 15: recording unit, 16: output unit, 20: input device, 30: display device

Claims

[Claim 1] A decision tree discrimination support device that supports analysis by decision tree discrimination of a whole data group having a plurality of analysis data including attribute data on attributes, values ​​of a plurality of adopted variables, and suitability data of suitability or insuitability as a result of discrimination, comprising: a storage unit in which analysis data of all the data groups is stored; an extracting unit that extracts a corresponding data group corresponding to attribute data selected via an input device from the entire stored data group; a setting unit that sets a decision tree for the corresponding data group via the input device from the values ​​of the adopted variables and threshold values ​​of the adopted variables that serve as stratification conditions, and also sets a prediction result of whether the branch determined by the decision tree is appropriate or inappropriate; an accuracy calculation unit that calculates an analysis accuracy consisting of a precision rate and a recall rate of the decision tree from the suitability data of the corresponding data group and the prediction result set by the setting unit; a recording unit that, for each change of the setting by the setting unit, accumulatively records the stratification conditions and the analysis accuracy of the branch having the largest number of pieces of analysis data predicted to be appropriate among the branches for each of the decision trees; and an output unit that outputs the recorded branch classification conditions and the analysis accuracy to a display device.

Images