Method for outputting causal relation calculation result and information processing device

Abstract

To improve the legibility of a causal relation map by accurately extracting and outputting a variable with high importance to an objective variable from the causal relation map showing respective causal relations among a plurality of variables.SOLUTION: An information processing device calculates first importance being importance of respective variables to an objective variable on the basis of the strength of a causal relation between respective variables from the objective variable till reaching the respective variables from the objective variable when tracing back to a variable existing on a cause side closer than a prescribed objective variable from the objective variable designated in the causal relation map showing the respective variables among a plurality of variables, and outputs an arithmetic result of a causal relation by extracting one or more variables whose importance to the objective variable is higher than prescribed reference on the basis of the first importance of the respective variables.SELECTED DRAWING: Figure 6

Description

[Technical field]

[0001] The present disclosure relates to a method for outputting a result of calculation of a causal relationship and an information processing device. [Background technology]

[0002] In recent years, artificial intelligence (AI) has been used in various fields. One of the techniques that AI uses is the technique of analyzing causal relationships between different elements.

[0003] For example, Patent Document 1 discloses an estimation device that estimates the amount of influence of an explanatory variable on a dependent variable. Specifically, the estimation device of Patent Document 1 includes a data acquisition unit that acquires correspondence data indicating the correspondence between a plurality of explanatory variables and a dependent variable, a partial correlation coefficient calculation unit that calculates a partial correlation coefficient between the explanatory variables and the dependent variable from the acquired correspondence data, a variable selection unit that selects a specific explanatory variable from among the plurality of explanatory variables according to the calculated partial correlation coefficient, a diagram creation unit that performs a covariance structure analysis on the explanatory variables selected by the variable selection unit and creates a causal diagram using the analysis result, and an influence amount estimation unit that estimates the amount of influence of the explanatory variables on the dependent variable using the created causal diagram.

[0004] According to the estimation device of Patent Document 1, by selecting a specific explanatory variable having a large influence on the dependent variable from among multiple explanatory variables according to the partial correlation coefficient between the multiple explanatory variables and the dependent variable, even if some explanatory variables with no influence are deleted from the multiple explanatory variables, the causal relationship toward the dependent variable will not be affected, or even if they have an influence, the influence can be treated as the influence of the selected explanatory variable without being deleted, thereby making it possible to suppress a decrease in the estimation accuracy of the influence of the explanatory variables on the dependent variable. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] JP 2020-135591 A Summary of the Invention [Problem to be solved by the invention]

[0006] However, in the estimation device of Patent Document 1, there is a risk that important variables that have an indirect causal relationship with the objective variable are deleted. Specifically, the estimation device of Patent Document 1 has a configuration in which, before analyzing the causal relationship, partial correlations between multiple explanatory variables and the objective variable are calculated by sparse modeling, and a specific explanatory variable that has a large influence on the objective variable is selected. It is considered that the explanatory variables deleted by this method include variables that have no direct relationship with the objective variable but are related to the objective variable on the upstream side away from the objective variable when tracing back upstream from the objective variable. In the estimation device of Patent Document 1, even if a variable is related to the objective variable, a variable that has been omitted from the selection of explanatory variables is not reflected in the final result, and there is a risk that the accuracy of the influence of the explanatory variable on the objective variable is not guaranteed.

[0007] The present disclosure has been made in consideration of the above circumstances, and an object of the present disclosure is to provide a method for outputting causal relationship calculation results and an information processing device that are capable of accurately extracting and outputting variables that are highly important to a target variable from a causal relationship map that shows the causal relationships between a plurality of variables, thereby improving the readability of the causal relationship map. [Means for solving the problem]

[0008] In order to solve the above problem, according to an aspect of the present disclosure, there is provided a method for outputting a causal relationship calculation result that outputs a result of estimating a causal relationship between a plurality of variables, the method including: a first importance calculation process that calculates a first importance, which is the importance of each variable with respect to the objective variable, based on the strength of the causal relationship between each variable from a predetermined objective variable specified in a causal relationship map indicating the causal relationships between the plurality of variables, when tracing back from the objective variable to each of the variables that are on the causal side of the objective variable; and an output process that extracts one or more variables whose importance with respect to the objective variable is higher than a predetermined standard, based on the first importance of each of the variables, and outputs a calculation result of the causal relationship.

[0009] Moreover, in order to solve the above problem, according to another aspect of the present disclosure, there is provided an information processing device that estimates a causal relationship between a plurality of variables, the information processing device including one or more processors and one or more memories communicably connected to the one or more processors, wherein the one or more processors execute a first importance calculation process that calculates a first importance, which is the importance of each variable with respect to the objective variable, based on the strength of the causal relationship between each variable from a predetermined objective variable specified in a causal relationship map indicating each causal relationship between the plurality of variables to each variable that is on the causal side of the objective variable, and an output process that extracts one or more variables whose importance with respect to the objective variable is higher than a predetermined standard, based on the first importance of each of the variables, and outputs a calculation result of the causal relationship. Effect of the Invention

[0010] As described above, according to the present disclosure, it is possible to accurately extract and output variables that are highly important to a target variable from a causal relationship map that shows the causal relationships between multiple variables, thereby improving the readability of the causal relationship map. [Brief description of the drawings]

[0011] [Figure 1] This is an explanatory diagram showing an example of a causal relationship map in which causal relationships are calculated using artificial intelligence AI. [Diagram 2] FIG. 11 is an explanatory diagram showing an example of a causal relationship map reconstructed by extracting only variables whose strength of causal relationship between variables is equal to or greater than a threshold value. [Diagram 3] 3 is an explanatory diagram showing an example of a causal relationship map reconstructed by extracting only variables whose strength of causal relationships between variables is equal to or greater than a threshold lower than the threshold in FIG. 2. FIG. [Figure 4] 1 is a block diagram functionally illustrating a configuration example of an information processing device according to an embodiment of the present invention. [Diagram 5] 10 is a flowchart of a causal relationship output process by the information processing device according to the present embodiment. [Figure 6] 4 is an explanatory diagram showing a process performed by a condition setting unit of the information processing device according to the embodiment; FIG. [Figure 7] 4 is an explanatory diagram showing a method for calculating the importance of each variable with respect to a target variable by a first importance calculation unit of the information processing device according to the present embodiment. FIG. [Figure 8] 10 is an explanatory diagram showing a result of calculating the importance of each variable with respect to a target variable by the information processing device according to the present embodiment. FIG. [Figure 9] 4 is an explanatory diagram showing a method for calculating the importance of each variable with respect to a target variable by a second importance calculation unit of the information processing device according to the present embodiment. FIG. [Figure 10] 11 is an explanatory diagram showing a result of calculating the importance of each variable with respect to an underlying cause by the information processing device according to the embodiment; FIG. [Figure 11] 11 is an explanatory diagram showing an example of a causal relationship map reconstructed by an output processing unit of the information processing device according to the embodiment; FIG. [Figure 12] FIG. 13 is an explanatory diagram showing the results of calculating the importance (first importance) of each variable with respect to the objective variable, with the threshold value of the strength of the causal relationship set to 0.32. [Figure 13] FIG. 13 is an explanatory diagram showing the results of calculating the importance (first importance) of each variable with respect to the objective variable, with the threshold value of the strength of the causal relationship set to 0.20. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0012] Hereinafter, preferred embodiments of the present disclosure will be described in detail with reference to the accompanying drawings. In this specification and the drawings, components having substantially the same functional configurations are denoted by the same reference numerals, and redundant description will be omitted.

[0013] <1. Overview of the technology disclosed herein> First, an overview of the technology of the present disclosure will be described.

[0014] Causal relationship maps output by inputting data into artificial intelligence (AI) become less readable when the number of variables becomes huge. For example, in the process of developing a continuously variable transmission (CVT) for an automobile, if the sound transmitted from the continuously variable transmission into the passenger compartment becomes a problem, it is necessary to suppress that sound. In that case, a causal relationship analysis is required to analyze which element of the continuously variable transmission is the origin of the sound, and which element the sound propagates through to generate in the passenger compartment. When analyzing the causal relationship between the sound generated by the continuously variable transmission and the sound inside the passenger compartment, the elements that must be considered are huge and complex, and the number of measurement points for measuring the data used in the analysis also becomes huge.

[0015] Below, some examples of measurement points to be used when analyzing the causal relationship between the sound generated by the continuously variable transmission and the sound inside the vehicle cabin are listed. (Hydraulic pressure of hydraulic circuit of continuously variable transmission [MPa]) -Oil pressure at the oil intake of the oil pump -Oil pressure at the oil discharge section of the oil pump - Hydraulic pressure in the main chamber hydraulic circuit of the control valve - Hydraulic pressure in the hydraulic circuit in the auxiliary chamber of the control valve -others

[0016] (Transmission case vibration [m / s 2 ]) -Oil pump case vibration - Vibration of the converter case -Extension case vibration -others

[0017] (Sound [Pa]) -Noise near the continuously variable transmission -Sound from the driver's seat -Sound from the passenger seat -Sound in the back seat -others

[0018] Since the structure of a continuously variable transmission is complex, the factors related to the hydraulic pressure of the hydraulic circuit of the continuously variable transmission and the vibration of the transmission case are numerous and complex, including factors other than those exemplified above, and the number of measurement points is also enormous. For example, if the objective variable is the sound in the driver's seat, the factors (variables) with low importance to the objective variable are the sound in the back seat or the vibration of the extension case, and the factors (variables) that are the root cause include the hydraulic pressure of the oil suction part of the oil pump. A causal relationship map, which is the output result of calculating causal relationships using artificial intelligence (AI), shows the causal relationships for all variables, including those with low importance, and therefore tends to be difficult to read.

[0019] Figure 1 shows a causal relationship map that is the output result of calculating causal relationships using artificial intelligence (AI) with 40 pieces of measurement data. In a causal relationship map, each variable x with a causal relationship is linked from "cause (parent) to result (child)," and many directed lines (arrows) connecting variables with causal relationships are drawn in a complex manner. In contrast, if it were possible to extract and show only the variables with high importance to the target variable, the readability of the calculation results of causal relationships would be improved.

[0020] Figure 2 shows an example of a causal relationship map reconstructed by extracting only variables x whose strength of causal relationship (degree of relationship) b between variables is 0.32 or more. The causal relationship map shown in Figure 2 has improved readability of the calculation results of causal relationships because variables with low importance to the target variable have been deleted compared to the causal relationship map shown in Figure 1. In this way, the readability of the calculation results of causal relationships can be improved by setting a threshold value for the strength b of the causal relationship between variables and reconstructing the causal relationship map by extracting only variables x that satisfy the condition.

[0021] However, while a causal relationship map shows the strength b of the causal relationship between each variable (cause variable and result variable) that has a causal relationship, it does not show the strength of the causal relationship of each variable x with a certain objective variable. For this reason, it is not easy to narrow down the variables that are important to the objective variable. In addition, when setting a threshold value for the strength b of the causal relationship between variables and extracting some variables x to reconstruct the causal relationship map, the readability can change in many ways depending on the threshold setting.

[0022] Figure 3 shows an example of a causal relationship map reconstructed by extracting only variables x whose causal relationship strength (degree of relationship) b between variables is 0.20 or more. In the example shown in Figure 3, more variables are extracted than in the example shown in Figure 2 with a threshold of 0.32, and the readability has changed. In addition, lowering the threshold results in multiple paths leading from multiple source causes to the target variable, which is not a desirable result.

[0023] In light of this background, the present disclosure aims to improve the readability of the causal relationship analysis results by accurately extracting variables x that are highly important to a target variable and outputting only the importance of the highly important variables x.

[0024] <2. Configuration of information processing device> Next, a configuration example of an information processing apparatus capable of executing the method for outputting a causal relationship calculation result according to an embodiment of the present disclosure will be described.

[0025] FIG. 4 is a block diagram showing a functional configuration example of the information processing device 10. As shown in FIG.

[0026] The information processing device 10 functions as a device that outputs the analysis result of causal relationships by executing a computer program by one or more processors such as a CPU (Central Processing Unit). The computer program is a computer program for causing the processor to execute the operation to be executed by the information processing device 10, which will be described later. The computer program executed by the processor may be recorded on a recording medium that functions as a storage unit (memory) 19 provided in the information processing device 10, or may be recorded on a recording medium built into the information processing device 10 or any recording medium that can be externally attached to the information processing device 10.

[0027] Examples of recording media for recording computer programs include magnetic media such as hard disks, floppy disks, and magnetic tapes; optical recording media such as CD-ROMs (Compact Disk Read Only Memory), DVDs (Digital Versatile Disks), and Blu-ray (registered trademark); magneto-optical media such as floptical disks; memory elements such as RAMs (Random Access Memory) and ROMs (Read Only Memory); and flash memories such as USB (Universal Serial Bus) memories and SSDs (Solid State Drives), as well as other media capable of storing programs.

[0028] The information processing device 10 includes a communication unit 11, an input unit 13, an output unit 15, a causal relationship analysis (artificial intelligence) AI 17, a storage unit 19, and a processing unit 20. The processing unit 20 is configured to include one or more processors such as a CPU. The processing unit 20 executes a computer program stored in the storage unit 19 to perform a calculation for outputting an analysis result of the causal relationship, and outputs the calculation result.

[0029] The communication unit 11 is an interface for acquiring a plurality of pieces of measurement data used in the causal relationship analysis process. For example, the communication unit 11 has a connector to which a storage device storing the measurement data or a computer is connected.

[0030] The input unit 13 is a means for a user to input operations for the information processing device 10. For example, the input unit 13 may be a mouse or a keyboard, but may be any other input means.

[0031] The output unit 15 outputs an operation screen and calculation results. For example, the output unit 15 includes at least one of an image display device and a printer.

[0032] The causal relationship analysis AI 17 is a storage device that stores an artificial intelligence program for generating a causal relationship map based on input measurement data.

[0033] The storage unit 19 includes one or more memories, and stores computer programs executed by the processing unit 20, various parameters used in the arithmetic processing, and information on the results of the calculations. The storage unit 19 may be a magnetic medium such as a hard disk, a floppy disk, or a magnetic tape, an optical recording medium such as a CD-ROM, a DVD, or a Blu-ray (registered trademark), a magneto-optical medium such as a floptical disk, a memory element such as a RAM, a flash memory such as a USB memory or an SSD, or other recording medium. A part of the storage unit 19 is used as a work area for the processing unit 20.

[0034] The processing unit 20 includes one or more processors, and includes an acquisition unit 21, a causal relationship map generation processing unit 22, a condition setting unit 23, a first importance calculation unit 24, a second importance calculation unit 25, and an output processing unit 26. The functions of these units are realized by the execution of a computer program by the processor. Note that some of the acquisition unit 21, the causal relationship map generation processing unit 22, the condition setting unit 23, the first importance calculation unit 24, the second importance calculation unit 25, and the output processing unit 26 may be configured by hardware such as an analog circuit.

[0035] The acquisition unit 21 acquires a plurality of pieces of measurement data stored in a storage device or a computer connected to the communication unit 11 .

[0036] The causal relationship map generation processing unit 22 inputs a plurality of measurement data to the causal relationship analysis AI, and generates a causal relationship map that indicates the strength b of the causal relationship between each variable included in the measurement data. Note that, when using data of a causal relationship map generated in advance, the information processing device 10 does not need to include the causal relationship map generation processing unit 22 and the causal relationship analysis AI 17.

[0037] The condition setting unit 23 performs a condition setting process to set conditions used in a calculation process for extracting a variable x with high importance from multiple variables x included in the causal relationship map. Specifically, when tracing back from a specified target variable to a variable x on the cause side, the condition setting unit 23 selects a variable x with the maximum strength b of the causal relationship and sets a main path back to the variable x assumed to be the original cause. In addition, the condition setting unit 23 sets the value of the minimum strength b of the causal relationship among the strengths b of the causal relationship between each variable present in the set main path as a threshold value.

[0038] The first importance calculation unit 24 performs a first importance calculation process to calculate the first importance, which is the importance of each variable x with respect to the objective variable, based on the strength b of the causal relationship between each variable from the objective variable to each variable x when tracing back from the objective variable to the variable x that is on the cause side of the objective variable. In this embodiment, the first importance calculation unit 24 selects variables x whose causal relationship strength is equal to or greater than the threshold set by the condition setting unit 23 from the objective variable, and traces back to the variable x on the cause side to calculate the first importance of each variable x with respect to the objective variable.

[0039] The second importance calculation unit 25 performs a second importance calculation process in which, starting from the variable x assumed to be the original cause from the result of the processing by the condition setting unit 23, the second importance calculation process calculates the second importance, which is the importance of each variable x with respect to the variable x assumed to be the original cause, based on the strength b of the causal relationship between each variable x from the variable x assumed to be the original cause to each variable x when tracing each variable x on the result side while selecting variables x having a causal relationship strength b equal to or greater than the threshold value set by the condition setting unit 23.

[0040] The output processing unit 26 extracts one or more variables x whose importance to the objective variable is higher than a predetermined standard based on the first importance of each variable, and outputs the calculation result of the causal relationship. In this embodiment, the output processing unit 26 extracts one or more variables x whose first importance and second importance are both positive values, and outputs the calculation result of the causal relationship.

[0041] <3. Operation of the information processing device> So far, we have described an example of the configuration of the information processing device 10. Next, we will specifically describe the operation of the causal relationship output process performed by the information processing device 10.

[0042] FIG. 5 shows a flowchart of the causal relationship output process by the information processing device 10. First, the causal relationship map generation processing unit 22 inputs a plurality of measurement data to the causal relationship analysis AI 17 to generate a causal relationship map (step S11). The process of generating a causal relationship map using the causal relationship analysis AI 17 may be performed by a known technique. In addition, when data of a causal relationship map generated in advance is used, step S11 may be omitted.

[0043] Next, the condition setting unit 23 selects a variable x with the maximum causal strength b from the specified objective variable as a starting point, traces back to the most upstream, sets the most upstream variable x as the original cause, and sets the traced path as the main path (step S13). Next, the condition setting unit 23 sets the minimum causal strength b in the main path as a threshold (step S15).

[0044] 6 is an explanatory diagram showing the processing by the condition setting unit 23. Here, for ease of understanding, the processing by the condition setting unit 23 will be explained using a causal relationship map of six variables x1 to x6.

[0045] In the causal relationship map shown in Fig. 6, it is assumed that the variable x6 is the objective variable. First, the condition setting unit 23 compares the strengths b42 and b43 of the causal relationship between the two variables x2 and x3 on the cause side having a causal relationship with the objective variable x6 on the result side. In the example shown in Fig. 6, the strength b42 of the causal relationship between the variable x2 and the objective variable x4 is 0.8, and the strength b43 of the causal relationship between the variable x3 and the objective variable x4 is 0.6, so the condition setting unit 23 selects the variable x2.

[0046] Since the variable x on the cause side having a causal relationship with the selected variable x2 on the result side is only variable x1, the condition setting unit 23 selects variable x1. Since there is no variable having a causal relationship with variable x1 on the result side, the condition setting unit 23 sets the most upstream variable x1 as the original cause. In addition, the path going back from the objective variable x4 to the variable x1, which is the original cause, is set as the main path.

[0047] Furthermore, the condition setting unit 23 sets the minimum strength of causal relationship among the strengths b of causal relationships between variables in the set main path as the threshold. In the example shown in Fig. 6, the strength b21 of the causal relationship between the variable x1, which is the original cause, and the variable x2 is 0.4, and the strength b42 of the causal relationship between the variable x2 and the objective variable x4 is 0.8, so the condition setting unit 23 sets 0.4 as the threshold.

[0048] Next, the first importance calculation unit 24 starts from the objective variable x4, traces back through variables x whose causal relationship strength b is equal to or greater than the threshold value set by the condition setting unit 23, and calculates the first importance, which is the importance of each variable x with respect to the objective variable x4 (step S17).

[0049] Figure 7 shows a method for calculating the importance of each variable x with respect to the objective variable x4 in the causal relationship map shown in Figure 6. The importance of variables x2 and x3 that have a direct causal relationship with the objective variable x4 as the result side is calculated by the strength of the causal relationship between the variables x2 and x3 with respect to the objective variable x4. The importance of variable x5 that is connected to the objective variable x4 by only one path is calculated by multiplying the strengths of the causal relationships b43 and b35 between the variables from the objective variable x4 to the variable x5.

[0050] The importance of the variable x1, which is connected to the objective variable x4 through multiple paths (two in FIG. 7), is calculated by adding the value (b42×b21) obtained by multiplying the strengths b42 and b21 of the causal relationships between the variables on the path (x4-x2-x1) from the objective variable x4 to the variable x1, and the value (b43×b31) obtained by multiplying the strengths b43 and b31 of the causal relationships between the variables on the other path (x4-x3-x1). At this time, if the strength b of the causal relationship is less than the threshold, the strength b of the causal relationship is calculated as zero. Note that the importance of the objective variable x4 itself and the importance of the variable x6, which cannot be traced back from the objective variable x4, cannot be calculated (zero).

[0051] 8 shows the results of calculating the importance of each variable x with respect to objective variable x4 when the threshold value is set to 0.4. The importance of variable x1 with respect to objective variable x4 is 0.32 (=0.8×0.4+0.6×0), the importance of variable x2 with respect to objective variable x4 is 0.8, the importance of variable x3 with respect to objective variable x4 is 0.6, the importance of variable x4 with respect to objective variable x4 is 0, the importance of variable x5 with respect to objective variable x4 is 0.24 (=0.6×0.4), and the importance of variable x6 with respect to objective variable x4 is 0.

[0052] When calculating the importance (first importance) of each variable x with respect to the objective variable x4, the calculation result will differ depending on the setting of the threshold value for the strength b of the causal relationship.

[0053] Next, the second importance calculation unit 25 starts from the original cause of the main path, traces the variables x with the strength b of the causal relationship equal to or greater than a threshold, and calculates the second importance of each variable x with respect to the original cause variable x1 (step S19). Here, starting from the original cause variable x1, the second importance of each variable x with respect to the original cause variable x1 is calculated in the same manner as in step S17.

[0054] Figure 9 shows how to calculate the importance of each variable x with respect to the variable x1, which is the original cause, in the causal relationship map shown in Figure 6. The importance of variables x2 and x3, which have a direct causal relationship with variable x1 as the cause, is calculated by the strength of the causal relationship between variables x2 and x3 with variable x1. The importance of variable x6, which is connected to variable x1 by only one path, is calculated by multiplying b21 and b62, the strengths of the causal relationships between each variable from variable x1 to variable x6.

[0055] The importance of the objective variable x4, which is connected to the variable x1 through multiple paths (two in FIG. 9), is calculated by adding the value (b21×b42) obtained by multiplying the strengths b21 and b42 of the causal relationships between the variables on the path (x1-x2-x4) from the variable x1 to the objective variable x4 by the strengths b31 and b43 of the causal relationships between the variables on other paths (x1-x3-x4). In this case, if the strength b of the causal relationship is less than a threshold, the strength b of the causal relationship is calculated as zero. Note that the importance of the variable x1 itself and the importance of the variable x5, which cannot be traced from the variable x1 in the direction of the directed line, cannot be calculated (zero).

[0056] Figure 10 shows the results of calculating the importance of each variable x with respect to variable x1 when the threshold value is 0.4. The importance of objective variable x4 with respect to variable x1 is 0, the importance of variable x2 with respect to variable x1 is 0.4, the importance of variable x3 with respect to variable x1 is 0.3, the importance of objective variable x4 with respect to variable x1 is 0.32 (=0.4×0.8+0×0.6), the importance of variable x5 with respect to variable x1 is 0, and the importance of objective variable x6 with respect to variable x1 is 0.28 (=0.4×0.7).

[0057] Next, the output processing unit 26 extracts only variables x whose first importance calculated in step S17 and second importance calculated in step S19 are both greater than zero, and outputs the calculation result of the causal relationship (step S21). That is, the output processing unit 26 extracts variables x whose first importance with respect to the objective variable x4 and second importance with respect to the variable x1, which is the original cause, are each positive, reconstructs the causal relationship map, and causes the output unit 15 to output the map. The output processing unit 26 may output graphs of the first importance and the second importance shown in Figs. 8 and 10 together with the reconstructed causal relationship map. This allows the user to grasp the importance of each variable with respect to the objective variable or the variable, which is the original cause.

[0058] Fig. 11 is an explanatory diagram showing an example of a reconstructed causal relationship map. The causal relationship map shown in Fig. 11 is a causal relationship map in which a variable x whose first importance and second importance are both positive values ​​is extracted from the causal relationship map shown in Fig. 6, and a variable x1 which is an original cause and a response variable x4 are connected by a directed line. That is, since the variable x whose first importance and second importance are both positive values ​​shown in Figs. 8 and 10 is only the variable x2, the output processing unit 26 reconstructs and outputs a causal relationship map by the variable x1 which is an original cause, the response variable x4, and the variable x2 whose importance to each of the variable x1 which is an original cause and the response variable x4 is indicated by a positive value.

[0059] In this way, according to the information processing device 10 of this embodiment, a causal relationship map is output in which variables x with high first importance to the objective variable x4 are extracted with high accuracy, instead of the strength b of the causal relationship between each variable. Therefore, even when the elements (variables) are huge and complex, the readability of the causal relationship map can be improved. Furthermore, according to the information processing device 10 of this embodiment, a causal relationship map is output in which variables x with high second importance to the variable x1, which is the root cause, as well as the first importance to the objective variable x4 are extracted, and the accuracy of the calculation result of the causal relationship can be guaranteed.

[0060] <4. Supplementary Information> The following provides additional information regarding the technology of the present disclosure. When a causal relationship map is reconstructed by extracting variables with high importance to a target variable using the method of causal relationship output processing according to this embodiment, the first importance and the second importance may differ depending on the setting of the threshold value for the strength b of the causal relationship.

[0061] 12 and 13 show the results of calculating the importance (first importance) of each variable x with respect to the objective variable x40 in the causal relationship map shown in FIG. 1 including 40 variables x1 to x40, with the variable x40 being the objective variable x40 and the thresholds of the strength b of the causal relationship being 0.32 and 0.2, respectively. As can be seen by comparing FIG. 12 and FIG. 13, the first importance of each variable x with respect to the objective variable x40 differs depending on the threshold of the strength of the causal relationship, and the number of variables x whose first importance is a positive value increases as the threshold is smaller. The same can be said about the second importance.

[0062] In contrast, in this embodiment, the condition setting unit 23 sets a main path from the objective variable x4 side back to the cause side, and the most upstream variable is set as the original cause, and the minimum value of the strength b of the causal relationship in the path is set as the threshold. Therefore, a variable x that is one original cause is specified, and the first importance calculation unit 24 can calculate the importance of the variable x that is connected to the objective variable x4 via the variables that are connected with a causal relationship strength b that is equal to or greater than the minimum value of the causal relationship strength b that connects the variables in the main path. In addition, the second importance calculation unit 25 can calculate the importance of the variable x that is connected from the variable x1 that is the original cause via the variables that are connected with a causal relationship strength b that is equal to or greater than the minimum value of the causal relationship strength b that connects the variables in the main path. Therefore, the threshold value to be set can be standardized based on the strength of the causal relationship in the main path, and the accuracy of extracting variables with high importance can be guaranteed to a certain standard.

[0063] Although the preferred embodiment of the present disclosure has been described in detail above with reference to the attached drawings, the technology of the present disclosure is not limited to such examples. It is clear that a person having ordinary knowledge in the technical field to which the present disclosure belongs can conceive of various modified or amended examples within the scope of the technical ideas described in the claims, and it is understood that these also naturally belong to the technical scope of the present disclosure. [Explanation of symbols]

[0064] 10: Information processing device 11: Communications Department 13: Input section 15: Output section 19: Storage section 20: Processing section 21: Acquisition section 22: Causal relationship map generation processing unit 23: Condition setting section 24: 1st importance calculation section 25:Second importance calculation section 26: Output processing section

Claims

1. A method for outputting a result of a calculation of a causal relationship, which outputs an analysis result of a causal relationship between a plurality of variables, By computer, a first importance calculation process for calculating a first importance, which is the importance of each variable with respect to the objective variable, based on the strength of the causal relationship between each variable from a predetermined objective variable specified in a causal relationship map indicating the causal relationships between each of the multiple variables, when tracing back from the objective variable to each of the variables that are on the cause side of the objective variable; an output process of extracting one or a plurality of variables whose importance to the objective variable is higher than a predetermined criterion based on the first importance of each of the variables, and outputting a calculation result of the causal relationship; A method for outputting the results of causal relationship calculations.

2. The computer includes: Before the first importance calculation process, a condition setting process is executed to trace back from the objective variable to the variables on the cause side, select a variable having the greatest strength of the causal relationship, trace back to the variable assumed to be the original cause, and set a main path to a threshold value that is the minimum strength of the causal relationship among the strengths of the causal relationships between the variables on the main path; In the first importance calculation process, while selecting variables having a causal relationship strength equal to or greater than the threshold value from the objective variable, the first importance of each variable with respect to the objective variable is calculated based on the strength of the causal relationship between each variable from the objective variable to each variable when tracing back to the variables on the cause side. The method for outputting a result of a causal relationship calculation according to claim 1 .

3. The computer includes: After the condition setting process and the first importance calculation process are executed, executing a second importance calculation process for calculating a second importance, which is the importance of each variable with respect to the variable assumed to be the original cause, based on the strength of the causal relationship between each variable from the variable assumed to be the original cause to each variable when tracing each variable on the result side while selecting variables having the strength of the causal relationship equal to or greater than the threshold value, starting from the variable assumed to be the original cause; In the output process, one or more variables for which the first importance and the second importance are both positive values ​​are extracted, and the calculation result of the causal relationship is output. The method for outputting a result of a causal relationship calculation according to claim 2 .

4. An information processing device that estimates a causal relationship between a plurality of variables, one or more processors; and one or more memories communicatively coupled to the one or more processors; the one or more processors: a first importance calculation process for calculating a first importance, which is the importance of each variable with respect to the objective variable, based on the strength of the causal relationship between each variable from a predetermined objective variable specified in a causal relationship map indicating the causal relationships between each of the multiple variables, when tracing back from the objective variable to each of the variables that are on the cause side of the objective variable; an output process of extracting one or a plurality of variables whose importance to the objective variable is higher than a predetermined criterion based on the first importance of each of the variables, and outputting a calculation result of the causal relationship; An information processing device that executes the above.

Images