Analysis device

The analysis device efficiently identifies and excludes outliers in multiple regression analysis, enhancing estimation accuracy by minimizing excluded data sets and maximizing the coefficient of determination.

JP2024164468AActive Publication Date: 2024-11-27TOYOTA JIDOSHA KK
View PDF 10 Cites 0 Cited by

Patent Information

Application Number
JP2023079960
Authority / Receiving Office
JP · JP
Patent Type
Applications
Current Assignee / Owner
Filing Date
2023-05-15
Publication Date
2024-11-27
Estimated Expiration
2043-05-15

AI Technical Summary

Technical Problem

Existing multiple regression analysis methods suffer from decreased estimation accuracy due to the inclusion of outliers, and manual outlier exclusion reduces efficiency.

Method used

An analysis device that calculates the degree of abnormality for each data set, sets thresholds to identify outliers, and performs multiple regression analysis on remaining data sets to minimize excluded data sets and maximize the coefficient of determination.

Benefits of technology

Enables efficient outlier exclusion, improving the accuracy of multiple regression analysis and preventing variations in estimation accuracy caused by manual outlier removal.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 2024164468000001_ABST
    Figure 2024164468000001_ABST
Patent Text Reader

Abstract

To provide an analysis device capable of efficiently excluding an outlier and accurately performing multiple regression analysis.SOLUTION: An analysis device 1 comprises: a calculation unit 11 which calculates an abnormality degree for each of a plurality of datasets composed of a plurality of explanatory variables and a plurality of objective variables; a setting unit 12 which sets a threshold value on abnormality degrees for identifying an exclusive dataset not subject to multiple regression analysis from the plurality of datasets; an analysis unit 13 which performs multiple regression analysis using the dataset left after the exclusive dataset identified on the basis of the threshold value among the plurality of datasets is excluded, and calculates a determination coefficient; a determination unit 14 which determines the minimization of the number of exclusive datasets and the maximization of the determination coefficient; and a regression equation acquisition unit which acquires a regression equation reflected with the results of the determination unit.SELECTED DRAWING: Figure 1
Need to check novelty before this filing date? Find Prior Art

Description

[Technical field]

[0001] The present disclosure relates to an analytical device. [Background technology]

[0002] There is a known multiple regression analysis method that expresses the relationship between a response variable and an explanatory variable as a relational equation and estimates the response variable using the relational equation. For example, Patent Document 1 discloses a multiple regression analysis device that divides multiple data sets into strata based on explanatory variables and thresholds for the explanatory variables, and performs multiple regression analysis on the divided data sets. [Prior art documents] [Patent documents]

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

[0004] When multiple data sets that are the subject of multiple regression analysis contain outliers, the estimation accuracy of the multiple regression analysis decreases. In order to perform multiple regression analysis with high accuracy, it is necessary to appropriately remove outliers from multiple data sets. However, when a user manually removes outliers, the efficiency of the work involved in the multiple regression analysis decreases. The technology disclosed in Patent Document 1 does not address such a problem.

[0005] The present disclosure makes it possible to efficiently remove outliers and perform multiple regression analysis with high accuracy. [Means for solving the problem]

[0006] The analysis apparatus according to the present disclosure includes a calculation unit that calculates a degree of anomaly for each of a plurality of data sets composed of a plurality of explanatory variables and a target variable; a setting unit that sets a threshold value for the degree of anomaly in order to identify an excluded data set from the plurality of data sets to be excluded from the multiple regression analysis; an analysis unit that performs the multiple regression analysis using a data set that remains after the excluded data set identified based on the threshold value is excluded from the plurality of data sets, and calculates a coefficient of determination; a determination unit that determines whether to minimize the number of the excluded data sets and maximize the coefficient of determination; and a regression equation acquisition unit that acquires a multiple regression equation that reflects the result of the determination unit. Effect of the Invention

[0007] The analysis device according to the present disclosure is capable of efficiently removing outliers and performing multiple regression analysis with high accuracy. [Brief description of the drawings]

[0008] [Figure 1] FIG. 1 is a block diagram showing a configuration of an analysis device according to an embodiment. [Diagram 2] FIG. 2 is a diagram showing a data group that is an example of a plurality of data sets. [Diagram 3] 4 is a flowchart showing a process performed by the analysis device according to the embodiment. [Figure 4] FIG. 4 is a diagram showing a specific example of a process performed by the analysis device according to the embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0009] Hereinafter, embodiments of the present disclosure will be described in detail with reference to the drawings. In each drawing, the same or corresponding elements are denoted by the same reference numerals. For clarity of explanation, duplicate explanations will be omitted as necessary.

[0010] (Configuration of analysis device 1) An analysis device 1 according to this embodiment will be described with reference to Fig. 1. Fig. 1 is a block diagram showing the configuration of the analysis device 1. The analysis device 1 includes a data acquisition unit 10, a calculation unit 11, a setting unit 12, an analysis unit 13, a determination unit 14, a regression equation acquisition unit 15, and a display control unit 16.

[0011] The analysis device 1 includes a processor, a memory, and a storage device, which are not shown in the figure. The storage device stores a program in which the processing according to this embodiment is implemented. The processor loads the program into the memory and executes the program, thereby realizing each function such as the data acquisition unit 10 described above. For example, a CPU (Central Processing Unit) may be used as the processor. A part or all of each functional component may be realized by dedicated hardware. Also, a part or all of each functional component may be realized by a plurality of information processing devices, circuits, etc. The plurality of information processing devices, circuits, etc. may be arranged in a centralized manner or in a distributed manner.

[0012] The data acquisition unit 10 acquires a plurality of data sets each composed of a plurality of explanatory variables and a target variable. The data acquisition unit 10 may acquire the plurality of data sets from a storage device (not shown) or from another device via a network (not shown).

[0013] Each of the multiple data sets (hereinafter referred to as "each data set") has multiple explanatory variables and a response variable. For example, each data set has M explanatory variables X1 to XM and one response variable Y. Here, "M" is an integer of 2 or more. The explanatory variables X1 to XM may be represented by a multidimensional vector. Furthermore, the data acquisition unit 10 acquires N such data sets. Here, "N" is an integer of 2 or more. The N data sets are associated with identification information for identifying each data set. The identification information is, for example, an ID assigned to each data set.

[0014] 2 is a diagram showing a data group G1 which is an example of a plurality of data sets acquired by the data acquisition unit 10. Here, an example of M=4 and N=30 is shown, but M and N are not limited to these. The data group G1 has data sets with ID=D01 to D30 in which IDs, explanatory variables X1 to X4, and a response variable Y are associated with each other. The degrees of anomaly shown in the rightmost column of the diagram are calculated by the calculation unit 11, which will be described later.

[0015] Returning to FIG. 1, the description will be continued. The calculation unit 11 calculates the degree of anomaly for each of a plurality of data sets acquired by the data acquisition unit 10. As described above, each data set has multidimensional explanatory variables X1 to X4. The calculation unit 11 calculates the multidimensional degree of anomaly of the explanatory variables of each data set. The calculation unit 11 may calculate the degree of anomaly using, for example, a technique such as Isolation Forest, which is a type of unsupervised learning. An example of the calculated degree of anomaly is shown in FIG. 2, as described above.

[0016] The setting unit 12 sets a threshold value for the degree of abnormality for identifying an excluded data set that is not to be subjected to the multiple regression analysis from among a plurality of data sets. The excluded data set corresponds to the outlier described above. The set threshold value is used in the analysis unit 13 to exclude the excluded data set from the plurality of data sets. The setting unit 12 may set the threshold value using a value set in advance. The setting unit 12 may set the threshold value using a value input by a user via an input unit (not shown). The setting unit 12 may also set an upper limit threshold and a lower limit threshold. For example, the setting unit 12 may set the upper limit threshold and the lower limit threshold based on at least one of the maximum value and the minimum value of the degree of abnormality. For example, the setting unit 12 may fix the upper limit threshold to the maximum value of the degree of abnormality and set the lower limit threshold to an arbitrary value.

[0017] Furthermore, when the determination unit 14 determines that the threshold does not satisfy a predetermined condition, the setting unit 12 sets a new threshold, thereby setting a new threshold different from the threshold that was previously set.

[0018] The analysis unit 13 performs multiple regression analysis using some or all of the multiple data sets acquired by the data acquisition unit 10. The analysis unit 13 first identifies an excluded data set from the multiple data sets to be excluded from the multiple data sets based on a threshold set by the setting unit 12. Next, the analysis unit 13 excludes the excluded data set from the multiple data sets. Then, the analysis unit 13 performs multiple regression analysis using the data set remaining after the excluded data set has been excluded from the multiple data sets, and calculates a coefficient of determination.

[0019] A specific description will be given using the example of the data group G1 described above. For example, it is assumed that the analysis unit 13 identifies three data sets as excluded data sets based on the threshold. The analysis unit 13 excludes the three excluded data sets from the data group G1, performs multiple regression analysis using the remaining 27 data sets, and calculates the coefficient of determination. Furthermore, when a new threshold is set in the setting unit 12, the analysis unit 13 performs multiple regression analysis using the newly set threshold. As a result, the analysis unit 13 performs multiple regression analysis on different data sets to be analyzed, and calculates the coefficient of determination corresponding to each data set.

[0020] The judgment unit 14 judges whether or not a predetermined condition is satisfied in the multiple regression analysis by the threshold set by the setting unit 12. The predetermined condition is that the number of excluded data sets is minimized and the coefficient of determination is maximized. The judgment unit 14 makes the above judgment using the threshold set by the setting unit 12 and the result of the multiple regression analysis by the analysis unit 13 using the threshold. As a result, the judgment unit 14 judges whether or not the number of excluded data sets is minimized and the coefficient of determination is maximized. In addition, the judgment unit 14 can judge whether or not the threshold satisfies the above predetermined condition. The judgment unit 14 identifies a threshold that satisfies the predetermined condition and outputs information on the threshold to the regression equation acquisition unit 15.

[0021] When the determination unit 14 determines that the threshold does not satisfy the predetermined condition, it causes the setting unit 12 and the analysis unit 13 to repeat the process. In this way, the determination unit 14 identifies a threshold that satisfies the predetermined condition by trial and error. In this way, the determination unit 14 realizes minimization of outliers and maximization of estimation accuracy of multiple regression analysis.

[0022] The regression equation acquisition unit 15 acquires a multiple regression equation reflecting the result of the determination unit 14. For example, the regression equation acquisition unit 15 acquires information on a threshold value that satisfies a predetermined condition. As a result, the regression equation acquisition unit 15 acquires a multiple regression equation obtained by multiple regression analysis using the threshold value. The regression equation acquisition unit 15 may output the acquired multiple regression equation to a display unit or the like via the display control unit 16. As a result, the analysis device 1 can present a multiple regression equation with improved estimation accuracy to a user.

[0023] The display control unit 16 causes a display unit (not shown) to display information related to the processing of the analysis device 1. The display information is, for example, but not limited to, the number of excluded data sets, the coefficient of determination, or the multiple regression equation acquired by the regression equation acquisition unit 15. The display control unit 16 may cause the display information to be displayed on an information terminal used by a user.

[0024] (Processing of analysis device 1) Next, the processing performed by the analysis device 1 will be described with reference to FIG. 3. FIG. 3 is a flowchart showing the processing performed by the analysis device 1. The overall flow of the processing will be described here, and specific examples of the processing will be described later. First, the data acquisition unit 10 acquires a plurality of data sets each composed of a plurality of explanatory variables and a target variable (S1). Next, the calculation unit 11 calculates the degree of anomaly for each data set (S2). Next, the setting unit 12 sets a threshold value for the degree of anomaly to identify an excluded data set (S3).

[0025] Next, the analysis unit 13 identifies an excluded dataset from the multiple datasets based on the set threshold (S4). The analysis unit 13 excludes the excluded dataset from the multiple datasets, performs multiple regression analysis using the dataset remaining after the excluded dataset has been excluded, and calculates a coefficient of determination (S5).

[0026] Next, the determination unit 14 determines whether or not the number of excluded data sets has been minimized and the coefficient of determination has been maximized (S6). If it is determined that the number of excluded data sets has been minimized and the coefficient of determination has been maximized (YES in S6), the regression equation acquisition unit 15 acquires a multiple regression equation reflecting the result of the determination unit 14 (S7). If it is determined that the number of excluded data sets has not been minimized and the coefficient of determination has not been maximized (NO in S6), the analysis device 1 returns to the process of step S3 and repeats the subsequent processes. The display control unit 16 may appropriately display display information related to the above-mentioned process on a display unit not shown.

[0027] The above-mentioned processing performed by the analysis device 1 will be specifically described with reference to Fig. 4. Fig. 4 is a diagram showing a specific example of processing performed by the analysis device 1. Here, it is assumed that the data acquisition unit 10 has acquired the data group G1 shown in Fig. 2. In Fig. 4, data groups G2 and G3 indicate data sets remaining after the excluded data sets identified based on the threshold are excluded from the data group G1.

[0028] The left side of Fig. 4 shows the data sets that each of the data groups G1 to G3 has. Unlike Fig. 2, the data sets are shown here in order from the smallest degree of anomaly to the largest. Also, explanatory variables X1 to X4 are not shown. The right side of Fig. 4 shows information related to multiple regression analysis using the data groups G1 to G3. The number of data sets N is the number of data sets acquired by the data acquisition unit 10. The number of data sets N is 30, and is common to the data groups G1 to G3.

[0029] The threshold Th is an example of a threshold set by the setting unit 12. Here, it is assumed that the setting unit 12 sets an upper threshold Th1 and a lower threshold Th2 as the threshold Th. The number of excluded data sets E is the number of data sets excluded from the data group G1. The number of excluded data sets E corresponds to the number of excluded data sets identified by the analysis unit 13 based on the thresholds set by the setting unit 12. The coefficient of determination R 2 is the coefficient of determination calculated by the analysis unit 13. The analysis device 1 minimizes the number of excluded data sets E and calculates the coefficient of determination R 2 A threshold Th that maximizes is calculated.

[0030] First, the setting unit 12 sets initial values ​​for the upper threshold Th1 and the lower threshold Th2. As shown in the area of ​​the data group G1 on the right side of the figure, the setting unit 12 sets the upper threshold Th1 and the lower threshold Th2 to 0.11 and -0.11, respectively. The data group G1 does not have any data set whose degree of abnormality is greater than the upper threshold Th1, and does not have any data set whose degree of abnormality is less than the lower threshold Th2. Therefore, the number of excluded data sets E is 0. The analysis unit 13 performs multiple regression analysis using all 30 data sets of the data group G1. Here, the analysis unit 13 calculates the coefficient of determination R 2 The result is 0.633.

[0031] Next, as shown in the area of ​​data group G2 on the right side of the figure, the setting unit 12 newly sets the upper threshold Th1 and the lower threshold Th2 to 0.11 and -0.05, respectively. The analysis unit 13 identifies the excluded data sets to be excluded from the data group G1 based on these thresholds. The analysis unit 13 identifies the data sets with IDs D07, D06, and D15 as the excluded data sets. Therefore, the number E of excluded data sets is 3. The analysis unit 13 performs multiple regression analysis using the data group G2 that remains after excluding these three data sets.

[0032] Here, the analysis unit 13 calculates the coefficient of determination R 2 The calculated value is 0.566. Compared to the data group G1, the coefficient of determination R2 The value of has become smaller. Therefore, the determination unit 14 determines that the set threshold does not satisfy the predetermined condition. The determination unit 14 causes the setting unit 12 and the analysis unit 13 to perform the process again.

[0033] Then, as shown in the area of ​​data group G3 on the right side of the figure, the setting unit 12 newly sets the upper threshold Th1 and the lower threshold Th2 to 0.11 and 0.01, respectively. Based on these thresholds, the analysis unit 13 identifies the data sets with IDs D07, D06, D15, . . ., D28, and D12 as excluded data sets. Therefore, the number E of excluded data sets is 10. The analysis unit 13 performs multiple regression analysis using the data group G3 that remains after excluding these 10 data sets.

[0034] Here, the analysis unit 13 calculates the coefficient of determination R 2 The calculated value is 0.694. Compared with the data groups G1 and G2, the coefficient of determination R 2 Therefore, the determination unit 14 determines whether the set threshold value is the value that minimizes the number of excluded data sets E and the coefficient of determination R 2 The determination unit 14 outputs information about the thresholds to the regression equation acquisition unit 15. The regression equation acquisition unit 15 acquires the information about the thresholds from the determination unit 14, and acquires a multiple regression equation in which the result of the determination unit 14 is reflected.

[0035] As described above, the analysis device 1 according to this embodiment determines whether the number of excluded data sets is minimized and the coefficient of determination is maximized, and obtains a multiple regression equation reflecting the result of the determination. In this way, the analysis device 1 can assist in removing outliers in multiple regression analysis. By using the analysis device 1, a user can efficiently improve the estimation accuracy of multiple regression analysis. Furthermore, the analysis device 1 can prevent variations in accuracy that occur when a user manually removes outliers.

[0036] The above-mentioned program includes a set of instructions (or software code) for making the computer perform the functions described in the embodiment when the program is loaded into the computer. The program may be stored in various types of non-transitory computer-readable media or tangible storage media. By way of example and not limitation, non-transitory computer-readable media or tangible storage media include random-access memory (RAM), read-only memory (ROM), flash memory, solid-state drive (SSD) or other memory technology, CD-ROM, digital versatile disc (DVD), Blu-ray® disk or other optical disk storage, magnetic cassette, magnetic tape, magnetic disk storage or other magnetic storage device. The program may also be transmitted on various types of transitory computer-readable media or communication media. By way of example and not limitation, transitory computer-readable media or communication media include electrical, optical, acoustic, or other forms of propagating signals.

[0037] The present disclosure is not limited to the above-described embodiment, and can be modified as appropriate without departing from the spirit and scope of the present disclosure. [Explanation of symbols]

[0038] 1: analysis device, 10: data acquisition unit, 11: calculation unit, 12: setting unit, 13: analysis unit, 14: judgment unit, 15: regression equation acquisition unit, 16: display control unit, E: number of excluded data sets, G1 to G3: data groups, N: number of data sets, R 2 : Coefficient of determination, Th: Threshold, Th1: Upper threshold, Th2: Lower threshold, X1 to X4 and XM: Explanatory variables, Y: Objective variable

Claims

[Claim 1] A calculation unit that calculates an anomaly degree for each of a plurality of data sets configured from a plurality of explanatory variables and a target variable; a setting unit that sets a threshold value related to the degree of anomaly in order to identify an excluded data set that is not to be subjected to multiple regression analysis from the plurality of data sets; an analysis unit that performs the multiple regression analysis using a dataset remaining after the excluded dataset identified based on the threshold is excluded from the plurality of datasets, and calculates a coefficient of determination; A determination unit that determines minimization of the number of excluded data sets and maximization of the coefficient of determination; a regression equation acquisition unit that acquires a multiple regression equation in which the result of the judgment unit is reflected, Analyzer.

Citation Information

Patent Citations

  • Monitoring method, monitoring program and monitoring system for production process

    JP2009037306A

  • Deterioration prediction mapping device and deterioration prediction mapping method

    JP2012013673A

  • Deterioration prediction mapping device and deterioration prediction mapping method for line hardware and the like

    JP2012013674A

  • Response surface model constructing device and program

    JP2012092760A

  • Power consumption analysis device and method for integrated circuit

    JP2012098952A