Method for displaying casual relationships and computer program

Abstract

To provide a technology for easily tracing an original cause from a displayed causal relationship image.SOLUTION: A method for displaying causal relationships of a plurality of variables includes the steps of: (a) generating a causal relationship image having nodes representing each of the plurality of variables including a plurality of feature amounts, extraction edges connecting between the nodes, the extraction edges representing edges satisfying predetermined display conditions, and an identification image for identifying a plurality of specific nodes corresponding to the plurality of feature amounts as one group among the plurality nodes; and (b) displaying the generated causal relationship image on a display device.SELECTED DRAWING: Figure 8

Description

[Technical field]

[0001] The present disclosure relates to techniques for displaying causal relationships among multiple variables. [Background technology]

[0002] Conventionally, as a technique for evaluating the causal relationship between multiple variables, a technique is known in which the presence or absence of a V-shaped structure is determined on a path between two non-independent variables, and if a conditional variable exists only on a path determined to have a V-shaped structure, a conditional independence test between two variables is not performed (Patent Document 1). A V-shaped structure is a graph structure in which a first variable and a second variable, which are independent, are not independent of a third variable. Conventionally, a technique is known for an industrial product manufactured through a complex process path, in which the mutual correlation strength between each variable in the manufacturing process history is calculated for all combinations, and manufacturing sequence information is used to generate a correlation structure model of the causal relationship between processes within the manufacturing line (Patent Document 2). Conventionally, an attempt has been made to use data acquired by a sensor of an injection molding machine to detect the tendency of a malfunction before it occurs and to suggest control parameters (Non-Patent Document 1). [Prior art documents] [Patent documents]

[0003] [Patent Document 1] JP 2013-206016 A [Patent Document 2] JP 2006-65598 A [Non-patent literature]

[0004] [Non-Patent Document 1] Toyota Industries Corporation (AI integration services), [online], FUJISOFT homepage, [searched on August 28, 2023], Internet < https: / / www.fsi.co.jp / cases / talk_shokki.html > Summary of the Invention [Problem to be solved by the invention]

[0005] When evaluating the causal relationship of multiple variables and displaying it on a display device, the multiple variables may include multiple feature quantities obtained from time series data. Examples of the multiple feature quantities obtained from time series data include maximum value, average value, number of peaks, overshoot amount, convergence time of overshoot, Manhattan distance and Euclidean distance indicating the difference from the target waveform data. In this way, many feature quantities are extracted from the time series data.

[0006] In the conventional technology, when the causal relationships between multiple feature quantities obtained from time-series data are evaluated and a causal relationship image that is the evaluation result is displayed on a display device, the number of edges indicating the causal relationships may become very large. In this case, it may be difficult to trace the original cause from the displayed causal relationship image. On the other hand, when the number of edges displayed on the display device is reduced using only the correlation strength as a threshold, it may be difficult to trace the original cause from the displayed causal relationship image. [Means for solving the problem]

[0007] According to a first aspect of the present disclosure, there is provided a method for displaying a causal relationship between a plurality of variables, the plurality of variables including a plurality of feature quantities obtained from at least one time-series data representing a physical quantity possessed by a single measurement target. The method includes: (a) generating a causal relationship image including a plurality of nodes representing the plurality of variables including the plurality of feature quantities, extraction edges connecting the nodes and satisfying a predetermined display condition, and an identification image for identifying a plurality of specific nodes corresponding to the plurality of feature quantities among the plurality of nodes as one collection; and (b) displaying the generated causal relationship image on a display device.

[0008] According to a second aspect of the present disclosure, there is provided a computer program for causing a computer to execute displaying a causal relationship between a plurality of variables on a display device. The plurality of variables include a plurality of feature quantities obtained from at least one time-series data representing a physical quantity obtained from a single measurement target. The computer program includes (a) a function for generating a causal relationship image having a plurality of nodes representing each of the plurality of variables including the plurality of feature quantities, extraction edges which are edges connecting the nodes and satisfy a predetermined display condition, and an identification image for identifying a plurality of specific nodes among the plurality of nodes corresponding to the plurality of feature quantities as one collection, and (b) a function for displaying the generated causal relationship image on a display device. [Brief description of the drawings]

[0009] [Figure 1] FIG. 1 is a block diagram of a data processing device according to a first embodiment. [Diagram 2] FIG. 13 is a diagram showing an example of time-series data. [Diagram 3] FIG. 4 is a diagram showing an example of a first setting image displayed on the display device. [Figure 4] FIG. 11 is a diagram showing an example of a second setting image displayed on the display device. [Diagram 5] 13 is an example of a third setting image displayed on the display device. [Figure 6] 11 is a flowchart showing a display process of a causal relationship image. [Figure 7] 7 is a detailed flowchart of step S20 in FIG. 6. [Figure 8] FIG. 13 is a diagram showing an example of a causal relationship image displayed on a display device. [Figure 9] FIG. 13 is a diagram for explaining a causal relationship image. [Figure 10] FIG. 13 is a diagram showing a drill-down display of a causal relationship image. [Figure 11] FIG. 13 is a diagram showing an example of a causal relationship image according to a reference example. [Figure 12] FIG. 11 is a diagram showing a causal relationship image according to the second embodiment. [Figure 13] FIG. 11 is a diagram showing a detailed image in the second embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0010] A. First embodiment: FIG. 1 is a block diagram of a data processing device 10 in the first embodiment. The data processing device 10 is a device that executes a method for analyzing and displaying the causal relationship between multiple variables. In a production process for manufacturing industrial products such as ink cartridges and circuit boards, the data processing device 10 estimates the causal relationship between variables for manufacturing the product, and generates and displays a causal relationship image IMG representing the estimated causal relationship. The causal relationship image IMG is, for example, an image using a graphical model using nodes ND representing variables and edges EG connecting the nodes ND. In this embodiment, the edges EG are directed graphs that are edges EG with arrows. The data processing device 10 may be composed of one device, or may be composed of two or more devices capable of data communication. In this embodiment, the data processing device 10 is an electronic calculator such as a personal computer.

[0011] The data processing device 10 includes a processor 20, a storage device 30, an input unit 40, and a display device 50. The processor 20 controls the operation of the data processing device 10 by executing various programs 32 stored in the storage device 30. Detailed functions of the processor 20 will be described later. The storage device 30 is composed of memories such as RAM and ROM. The storage device 70 stores various programs 32 for implementing each function of the processor 20, generation conditions 34 for generating a causal relationship image IMG, variable data 36, ​​and time-series data 38. The generation conditions 34 are stored in the storage device 30 in advance, or are stored by input from a user via the input unit 40. The generation conditions 34 include information for identifying a variable for which a causal relationship is to be estimated, among each variable of the variable data 36 described later.

[0012] The variable data 36 stores a plurality of variables that are candidates for estimating a causal relationship in association with a product identifier. The variable data 35 includes a plurality of variables used in a manufacturing process of the product and variables related to the product. The plurality of variables may be, for example, the temperature of a device for manufacturing the product, the room temperature in a factory for manufacturing the product, the time period for manufacturing the product, the dimensions of the product, or the mass of the product. The plurality of variables also includes a plurality of feature quantities obtained from at least one time-series data 38 that represents one physical quantity possessed by the same measurement target. The one physical quantity is pressure, temperature, current, voltage, or the like. The one physical quantity is acquired by a sensor. The one physical quantity may be represented by one time-series data 38, or may be represented by two or more time-series data 38.

[0013] For example, in a temperature sensor, the temperature, which is one physical quantity to be measured, is represented by one time series data 38. In a six-axis gyro sensor, the motion, which is one physical quantity to be measured, is represented by three time series data 38 for the acceleration of each of the X-axis, Y-axis, and Z-axis, and three time series data 38 for the angular velocity around the X-axis, Y-axis, and Z-axis. In a six-axis force sensor, the load, which is one physical quantity to be measured, is represented by three time series data 38 for the force of each of the X-axis, Y-axis, and Z-axis, and three time series data 38 for the torque around the X-axis, Y-axis, and Z-axis. As described above, one or more time series data 38 measured by one sensor can be said to be at least one time series data 38 representing one physical quantity possessed by the same measurement object. In addition, one physical quantity is a concept that includes not only a physical quantity represented in a single dimension such as temperature or pressure, but also a physical quantity represented in multiple dimensions such as motion or load. That is, in this embodiment, one physical quantity is obtained by detecting one physical event with a sensor.

[0014] The time series data 38 is time series data that represents a physical quantity detected by a sensor during a predetermined period of time. The time series data 38 is stored in the storage device 30 in association with a product ID.

[0015] FIG. 2 is a diagram showing an example of the time series data 38. The time series data 38 shown in FIG. 2 is data with a product ID of "0001". A plurality of feature values ​​Ft obtained from the time series data 38 will be described with reference to FIG. 2. FIG. 2 shows time series data 38 acquired by a temperature sensor when an unfinished product is heated for a time tx in a manufacturing process A, for example. A plurality of feature values ​​Ft are extracted from the acquired time series data 38. The extraction of the feature values ​​Ft may be performed by the data processing device 10 or by another device. The feature values ​​Ft of the plurality of feature values ​​Ft include a maximum value Ft_max, a minimum value Ft_min, an integral value Ft_int, and an average value Ft_ave. The feature values ​​Ft of the plurality of feature values ​​Ft include the number of peaks, the amount of overshoot, the convergence time of the overshoot, and the absolute value of the area of ​​the difference from the target waveform data. In addition, each of the multiple feature quantities Ft may be classification data indicating which cluster the time series data 38 of the target product ID is classified into when the time series data 38 for each product ID is clustered. For example, when the time series data 38 for each product ID is classified into four categories according to the quality of the product, such as "good product", "defective product 1", "defective product 2", and "defective product 3", the value according to each category becomes the classification data. For example, the classification data for "good product" has a value of "1", the classification data for "defective product 1" has a value of "2", the classification data for "defective product 2" has a value of "3", and the classification data for "defective product 3" has a value of "4". The clustering method is not limited to the above, and the K-means method or a method using a trained machine learning model may be used.

[0016] Returning to FIG. 1, the data processing device 10 will be further described. The input unit 40 is an interface that accepts information from the outside. The input unit 40 accepts input from a user, for example. The input unit 40 is, for example, a mouse, a keyboard, or a touch panel. The display device 50 displays various types of information. The display device 50 is, for example, a liquid crystal monitor.

[0017] The processor 20 executes various programs 32 in the storage device 30, thereby functioning as a generation unit 24 and a display control unit 28. The generation unit 24 generates a causal relationship image IMG using the generation conditions 34 stored in the storage device 30, and causes the display device 50 to display the generated causal relationship image IMG. The display control unit 28 controls the display device 50 to display various information on the display device 50. The display control unit 28 causes the display device 50 to display, for example, setting images IM1, IM2, and IM3 that accept input from the user for setting at least a part of the generation conditions 34, and the causal relationship image IMG. Details of the setting images IM1, IM2, and IM3 and the causal relationship image IMG will be described later. The setting image IM1 is also referred to as a first setting image IM1, the setting image IM2 is also referred to as a second setting image IM2, and the setting image IM3 is also referred to as a third setting image IM3.

[0018] The generating unit 24 includes a setting image generating unit 25, a causal relationship estimating unit 26, and a relationship image generating unit 27. The setting image generating unit 25 generates setting images IM1 and IM2 and displays them on the display device 50. The causal relationship estimating unit 26 estimates the causal relationship of a plurality of variables set as estimation targets. The causal relationship is estimated by calculating an index value indicating the strength of the relationship between two nodes ND and connecting the two nodes ND with an edge EG having a direction. The index value is, for example, a value using any of the following indices: mutual information, Pearson's product moment correlation, Hilbert-Schmidt independence criterion, partial correlation coefficient, correlation ratio, association coefficient, rank correlation, and entropy. In this embodiment, the causal relationship estimating unit 26 calculates the mutual information between two nodes ND to estimate the causal relationship.

[0019] The relationship image generating unit 27 generates, as a causal relationship image IMG, the causal relationship estimated by the causal relationship estimating unit 26. In this embodiment, the relationship image generating unit 27 generates an image using a graphical model as the causal relationship image IMG.

[0020] FIG. 3 is a diagram showing an example of a first setting image IM1 displayed on the display device 50. The first setting image IM1 is a screen for selecting a feature amount Ft, which is a variable for which a causal relationship is to be estimated, among each of the time series data 38. The first setting image IM1 includes an individual setting image IMI related to at least one of the time series data 38 representing one physical amount, and an instruction input image IMSa. The individual setting image IMI is provided for each of the time series data 38 representing one physical amount possessed by the same measurement target. In the example shown in FIG. 3, the individual setting image IMI includes a first individual setting image IMIa and a second individual setting image IMIb. The first individual setting image IMIa is an image for selecting a target for which a causal relationship is to be estimated from the feature amounts Ft_A0 to Ft_A5 extracted from the pressure data, which is the time series data 38. This pressure data is data acquired from a pressure sensor in the manufacturing process A. The second individual setting image IMIb is an image for selecting a target for inferring a causal relationship from among feature amounts B0 to B5 extracted from temperature data, which is time-series data 38. This temperature data is data acquired from a temperature sensor in manufacturing process B, which is executed after manufacturing process A.

[0021] The first individual setting image IMIa includes sample number information, which is the number of products from which pressure data was obtained, and text information, "pressure data," which is an explanation of the physical quantity represented by the time series data 38. The first individual setting image IMIa also includes explanatory information that explains the feature amounts Ft_A0 to Ft_A5, and check box images Bx formed next to each explanatory information. The explanatory information regarding the classification data by clustering the time series data 38 includes text information, "waveform cluster," cluster type, waveform data representing the representative time series data 38 for each cluster type, and numerical information, which is the number of products for each cluster type. The feature amount Ft_A5 is the distance, for example, Manhattan distance, between the average time series data obtained by averaging the time series data 38 of the cluster type with the largest number of classifications and the target time series data 38.

[0022] The second individual setting image IMIb includes sample number information, which is the number of products for which temperature data was obtained, and text information, "temperature data," which is an explanation of the physical quantity represented by the time series data 38. The second individual setting image IMIb also includes explanatory information that explains the feature amounts Ft_B0 to Ft_B5, and check box images Bx formed next to each explanatory information. The explanatory information regarding the classification data by clustering the time series data 38 includes text information, "waveform cluster," cluster type, waveform data representing the representative time series data 38 for each cluster type, and numerical information, which is the number of products for each cluster type. The feature amount Ft_B5 is the distance, for example, Manhattan distance, between the average time series data obtained by averaging the time series data 38 of the cluster type with the largest number of classifications and the target time series data 38.

[0023] The user selects the checkbox image Bx with a cursor or the like to check the checkbox image Bx, thereby selecting the checked feature amounts Ft_A0 to Ft_A5 and Ft_B0 to Ft_B5 as targets for which causal relationships are to be inferred.

[0024] For each of the first individual setting image IMIa and the second individual setting image IMIb, the feature amount Ft that is the subject of causal relationship estimation by checking the check box image Bx is displayed as a specific node NDf in the causal relationship image IMG. In this manner, among the multiple nodes ND, each of the multiple nodes ND corresponding to the multiple feature amounts Ft generated from one physical amount is called a specific node NDf. In this embodiment, the feature amount Ft that is the subject of causal relationship estimation among the feature amounts of the first time series data 38 is also called a first specific node NDf1, and the feature amount Ft that is the subject of causal relationship estimation among the feature amounts Ft of the second time series data 38 is also called a second specific node NDf2.

[0025] The instruction input image IMSa includes an OK image indicating confirmation and a cancel image indicating cancellation. When the user selects the OK image with the cursor or the like, the feature amounts Ft_A0 to Ft_A5 and Ft_B0 to Ft_B5 with the checkbox images Bx checked are determined as targets for which causal relationship estimation is performed. That is, the determined feature amounts Ft_A0 to Ft_A5 and Ft_B0 to Ft_B5 are stored in the generation conditions 34 of the storage device 30 in association with estimation target information indicating that they have been determined as targets for which causal relationship estimation is performed. In the example shown in FIG. 3, the feature amounts Ft_A0 to Ft_A3 and Ft_B0 to Ft_B3 are stored in the generation conditions 34 in association with the estimation target information. When the user selects the cancel image with the cursor or the like, the checkbox images Bx are reset to the initial state. For example, in the initial state, all the checkbox images Bx are checked.

[0026] 4 is a diagram showing an example of the second setting image IM2 displayed on the display device 50. The second setting image IM2 is an image for setting a temporal order for a plurality of variables. The direction of the arrow of the edge EG is determined by the temporal order. This will be described in detail later. The second setting image IM2 includes a variable selection image IM2a, a covariate selection image IM2b, an order setting image IMt, and an instruction input image IMSb.

[0027] The variable selection image IM2a includes individual images showing each variable that is a candidate for the estimation target of the causal relationship. The individual images are provided for each variable, or for each aggregate of multiple variables. As an example of treating multiple variables as one aggregate, there are multiple variables corresponding to multiple feature amounts Ft generated based on one physical quantity. For example, in the case of time series data 38 showing pressure data, as shown in FIG. 3, multiple variables such as feature amounts Ft_A0 to Ft_A5 are included. However, the feature amounts Ft_A0 to Ft_A5 are not displayed as individual images showing each variable, but are displayed on the display device 50 as individual images IMGr1 showing an aggregate called "first time series data (pressure)" in which the feature amounts Ft_A0 to Ft_A5 are treated as one aggregate. Also, for example, the feature amounts Ft_B0 to Ft_B5 extracted from the time series data 38 of temperature data are displayed on the display device 50 as individual images IMGr2 showing an aggregate called "second time series data (temperature)" instead of individual images showing each variable.

[0028] The covariate selection image IM2b includes a covariate image showing a covariate as a variable. In the example shown in Fig. 4, a room temperature image showing the room temperature in the factory where each manufacturing process is performed, a manufacturing time zone image showing the manufacturing time zone of the product, and a cooling water temperature image showing the temperature of the cooling water circulated in the manufacturing device that manufactures the product are illustrated as the covariate images.

[0029] The order setting image IMt has a plurality of frame-shaped images IMta to IMtd arranged in a horizontal line. These frame images can be newly added or deleted. The order of the frame-shaped images IMta to IMtd corresponds to the chronological order. As the frame-shaped images IMta to IMtd are arranged from the left to the right, they are downstream in the chronological order. The user moves the individual images and covariate images in the variable selection image IM2a and the covariate selection image IM2b by dragging them with the mouse or the like to move them into one of the frame-shaped images IMta to IMtd. In general, among the plurality of frame-shaped images IMta to IMtd, the objective variable is moved into the rightmost image IMtd, and the explanatory variables are moved into the other images IMta to IMtc. In the example shown in FIG. 4, the dimensions and mass of the product are moved into the image IMtd as the objective variables.

[0030] When the user performs the setting using the second setting image IM2, the user refers to the manufacturing sequence information for the product and moves the individual image or the covariate image into one of the frame-shaped images IMta to IMtd. The manufacturing sequence information is prepared in advance, and is, for example, a parts configuration table or manufacturing parts sequence information. The parts configuration table is information that hierarchically describes the configuration of parts from lower-level materials and parts to higher-level intermediate assemblies and final products, and represents the manufacturing sequence in the case of a product with a relatively simple configuration. The manufacturing parts sequence information may be created for products with complex manufacturing processes, and is information that collates the parts hierarchical configuration with the process sequence and adds incidental conditions for manufacturing.

[0031] The user refers to the manufacturing sequence information, and places, for example, an individual image showing a variable on a more upstream side among the individual images in the frame-shaped image IMta displayed further to the left. A plurality of variables extracted based on one physical quantity can be moved into any of the frame-shaped images IMta to IMtd as individual images IMGr1 and IMGr2 represented as a collection. This makes it possible to prevent a plurality of variables extracted based on one physical quantity from being placed in different frame-shaped images IMta to IMtd. The covariate image may be moved by the user into any of the frame-shaped images IMta to IMtd, or may be automatically moved into a predetermined frame-shaped image IMta to IMtd. The covariate may be automatically moved into the frame-shaped image IMta on the most upstream side. In addition, the covariate image may be placed in advance in any of the frame-shaped images IMta to IMtd as an initial setting.

[0032] Variables corresponding to the individual images and covariate images arranged in the frame-shaped images IMta to IMtd are set as nodes ND, and become targets of causal relationship estimation by the causal relationship estimation unit 26. Here, for the individual images IMGr1 and IMGr2 arranged in the frame-shaped images IMta to IMtd, the feature amount Ft, which is a variable for which the check box image Bx, input of which is accepted by the first setting image IM1 shown in FIG.

[0033] Furthermore, the causal relationship estimation unit 26 estimates the causal relationship according to a combination of the following first estimation algorithm and second estimation algorithm, or a combination of the second estimation algorithm and third estimation algorithm. <Estimation algorithm> First estimation algorithm: This method is generally classified as score-based, and is an algorithm that uses, for example, the Bayesian information criterion based on logarithmic likelihood as an index, and an empty graph with no edges is set as the initial state, and determines and greedily adds directed edges EG in the direction that improves the score while taking into account a penalty for complexity. During the search, prohibited edges can be excluded from the index calculation. In this case, the index value is calculated between variables arranged in adjacent frame-shaped images IMta to IMtd, and only the direction of the edge EG excluding the prohibited edge is calculated so that the variable in the frame-shaped image IMta to IMtd located on the left side of the adjacent frame-shaped images IMta to IMtd faces the variable in the frame-shaped image IMta to IMtd located on the right side. Second estimation algorithm: This method is generally classified as a constraint-based method, and is an algorithm that sets a graph structure in which all nodes are connected as the initial state, sequentially reduces undirected edges that indicate relationships using conditional independence, and uses a significance level as a threshold judgment to sequentially reduce edges EG, and determines a V structure that determines the direction of the edges EG by the presence or absence of conditional dependence using the significance level. Using the edge direction and chronological order according to the V structure, a result that matches the user's intention is obtained. In this case, an index value is calculated between variables arranged in adjacent frame-shaped images IMta to IMtd, and only the direction of the edges EG is calculated so that the variable in the frame-shaped images IMta to IMtd located on the left side of the adjacent frame-shaped images IMta to IMtd faces the variable in the frame-shaped images IMta to IMtd located on the right side, excluding prohibited edges. Third estimation algorithm: For each of the individual images IMGr1 and IMGr2, the direction of the arrow of the edge EG is determined according to a predetermined rule for the variables for which the causal relationship is to be estimated, which are set using the first setting image IM1 and the second setting image IM2. The predetermined rule is, for example, a rule that in the first setting image IM1 shown in Fig. 3, for variables arranged within the frame of the first individual setting image IMIa, edges EG between the variables do not exist in both directions within the same frame, that is, edges EG that directly connect nodes ND do not exist within the same frame.

[0034] The instruction input image IMSb includes an add image indicating an addition, a delete image indicating a delete, an OK image indicating an OK, and a cancel image indicating a cancel. When the user selects the add image with a cursor or the like, a new frame-shaped image is additionally displayed to the right of the currently displayed frame-shaped images IMta to IMtd. When the user selects at least one of the currently displayed frame-shaped images IMta to IMtd with a cursor or the like and then selects the delete image with a cursor or the like, the selected frame-shaped image IMta to IMtd is deleted. When the user selects the OK image with a cursor or the like, the variables arranged in the frame-shaped images IMta to IMtd are determined as targets for which a causal relationship is to be estimated, and are stored in the generation condition 34 of the storage device 30 in association with information indicating that the variables have been determined as targets for which a causal relationship is to be estimated and with position information of the frame-shaped images IMta to IMtd. When the user selects the cancel image with a cursor or the like, the second setting image IM2 is reset to the initial state. For example, in the initial state of the second setting image IM2, no individual images or covariate images are arranged within the frame-shaped images IMta to IMtd, but the individual images and covariate images are arranged within the frames of the variable selection and covariate selection.

[0035] 5 is an example of a third setting image IM3 displayed on the display device 50. The third setting image IM3 is a screen for setting predetermined display conditions of the edge EG. The third setting image IM3 has a first reference value setting image IM3a, a second reference value setting image IM3b, and an instruction input image IMSc.

[0036] The first reference value setting image IM3a is an image for setting a first display condition, which is a display condition of a specific edge EGf connecting two specific nodes NDf treated as one aggregate among a plurality of specific nodes NDf as a plurality of variables to be estimated included in each individual image IMGr1, IMGr2 displayed as an aggregate shown in FIG. 4. The first reference value setting image IM3a is an image for setting a first reference value, which is a threshold value for the absolute value of an index value indicating the strength of the relationship between two specific nodes NDf. The first reference value setting image IM3a has a first bar image IMra and a first value display image IMrb. The first bar image IMra can set the first reference value by moving the bar to the left and right by operating a cursor or the like by a user. The first value display image IMrb displays a numerical value according to the position of the bar of the first bar image IMra. Note that the first reference value may be set by inputting a numerical value into the first value display image IMrb. In this embodiment, mutual information is used as an index value indicating the strength of the relationship between two nodes ND, so the first reference value is a threshold value of mutual information. 1 extracts specific edges EGf having mutual information equal to or greater than a first reference value as extracted edges EGs from the calculated specific edges EGf. Then, the relationship image generating unit 27 generates the extracted edges EGs as a part of the causal relationship image IMG.

[0037] The second reference value setting image IM3b shown in FIG. 5 is an image for setting a second display condition, which is a display condition of other edges EGg different from the specific edge EGf among the edges EG connecting two nodes ND, which are variables set as the target of the estimation of the causal relationship. The second reference value setting image IM3b has a second bar image IMva and a second value display image IMvb. ​​The second bar image IMva can set a second reference value by moving the bar to the left or right by operating the cursor or the like by the user. The second value display image IMvb displays a numerical value according to the position of the bar of the second bar image IMva. Note that the second reference value may be set by inputting a numerical value into the second value display image IMvb. ​​In this embodiment, since the mutual information is used as an index value indicating the strength of the relationship between the nodes ND, the second reference value is a threshold value of the mutual information. The causal relationship estimation unit 26 shown in FIG. 1 extracts other edges EGg having mutual information equal to or greater than the second reference value as extracted edges EGs from among the calculated other edges EGg. Then, the relationship image generating unit 27 generates the extracted edges EGs as a part of the causal relationship image IMG.

[0038] Here, the second reference value is set to a value lower than the first reference value. For example, when a second reference value equal to or greater than the first reference value is set in the third setting image IM3, an error message may be displayed. In addition, the maximum value Smax2 that can be set for the second reference value may be set to a value less than the first reference value displayed in the first value display image IMrb so that a second reference value equal to or greater than the first reference value cannot be set.

[0039] The instruction input image IMSc includes an OK image indicating confirmation and a cancel image indicating cancellation. When the user selects the OK image with the cursor or the like, the first reference value and the second reference value displayed in the first value display image IMrb and the second value display image IMvb are stored in the generation condition 34 of the storage device 30 as display conditions.

[0040] As described above, the predetermined display conditions for extracting the extracted edges EGs to be generated as part of the causal relationship image IMG include a first display condition and a second display condition. The first display condition is a condition that the absolute value of the index value indicating the strength of the relationship between the nodes ND is equal to or greater than a first reference value. The second display condition is a condition that the absolute value of the index value is equal to or greater than a second reference value that is lower than the first reference value.

[0041] FIG. 6 is a flowchart showing a display process of the causal relationship image IMG. The display process is started when the data processing device 10 receives an instruction to start the display process inputted to the data processing device 10 by a user. First, in step S10, the causal relationship estimation unit 26 estimates a causal relationship for a plurality of variables for which a causal relationship is to be estimated. In step S10, the causal relationship estimation unit 26 identifies a plurality of variables for which a causal relationship is to be estimated by referring to the generation condition 34 and the variable data 36, ​​and calculates an index value and determines an edge EG including a direction for the plurality of variables for which a causal relationship is to be estimated according to the above-mentioned first estimation algorithm to third estimation algorithm. In this embodiment, the index value calculated in step S10 is mutual information. The mutual information is calculated using the following formula (1) for calculating the mutual information of two discrete random variables.

number

[0042] Next, in step S20, the relationship image generation unit 27 generates a causal relationship image IMG in accordance with the first display condition and the second display condition included in the generation condition 34. Next, in step S40, the display control unit 28 causes the display device 50 to display the causal relationship image IMG.

[0043] Fig. 7 is a detailed flowchart of step S20 in Fig. 6. First, in step S21, the relation image generating unit 27 selects one edge EG from the multiple edges EG determined in step S10. Next, in step S22, the relation image generating unit 27 determines whether the selected edge EG is a specific edge EGf connecting specific nodes NDf generated based on one physical quantity.

[0044] If it is determined in step S22 that the edge EG is a specific edge EGf, then in step S23 the relation image generating unit 27 determines whether or not the specific edge EGf satisfies a first display condition. That is, the relation image generating unit 27 determines that the specific edge EGf satisfies the first display condition when the absolute value of the index value between two specific nodes NDf connected by the specific edge EGf is equal to or greater than a first reference value.

[0045] If the determination in step S23 is "Yes", the relationship image generating unit 27 determines in step S24 to generate the specific edge EGf as an extracted edge EGs as an image constituting the causal relationship image IMG. On the other hand, if the determination in step S23 is "No", the relationship image generating unit 27 determines in step S25 not to generate the specific edge EGf determined in step S23 as an image constituting the causal relationship image IMG.

[0046] If it is determined in step S22 that the edge EG is not a specific edge EGf, then in step S26 the relation-image generating unit 27 determines whether or not another edge EGg, which is the edge EG determined not to be the specific edge EGf, satisfies a second display condition. That is, the relation-image generating unit 27 determines that the second display condition is satisfied when the absolute value of the index value between two nodes ND connected by the other edge EGg is equal to or greater than a second reference value.

[0047] If the determination in step S26 is "Yes", the relationship image generating unit 27 determines in step S27 to generate the other edge EGg as the extracted edge EGs as an image constituting the causal relationship image IMG. On the other hand, if the determination in step S26 is "No", the relationship image generating unit 27 determines in step S28 not to generate the other edge EGg determined in step S26 as an image constituting the causal relationship image IMG.

[0048] In step S29, the relationship image generating unit 27 judges whether or not the processing of steps S21 to S28 has been performed on all edges EG. If the judgment in step S29 is "No", the relationship image generating unit 27 executes step S21 again. In the re-executed step S21, one edge EG is selected from among the edge EGs for which the processing of steps S22 to S28 has not been performed up until the previous time. On the other hand, if the judgment in step S29 is "Yes", the relationship image generating unit 27 generates a causal relationship image IMG in step S30. In step S30, the relationship image generating unit 27 generates a causal relationship image IMG including at least the node ND, the extracted edge EGs, and an identification image IMf and a node explanation image, which will be described later.

[0049] FIG. 8 is a diagram showing an example of a causal relationship image IMG displayed on the display device 50. The causal relationship image IMG includes a node ND, an extracted edge EGs connecting two nodes ND, an identification image IMf, and an explanatory image explaining the node ND and the aggregate. The extracted edge EGs meets a predetermined display condition. The node ND in the causal relationship image IMG represents a plurality of variables set as the estimation target of the causal relationship using the first setting image IM1 shown in FIG. 3 and the second setting image IM2 shown in FIG. 4. The plurality of variables displayed as the node ND include a plurality of feature amounts Ft obtained from the time series data 38. The extracted edge EGs is an edge EG for which it is determined to generate an image in step S24 and step S27 in FIG. 7. The explanatory image is a text image explaining various images such as each node ND and the identification image IMf. As shown in FIG. 8, the text image is, for example, a "controlled variable a", a "controlled variable b", a "maximum value", and a "pressure data".

[0050] The identification image IMf is an image for identifying, as one aggregate, a plurality of specific nodes NDf corresponding to the feature amount Ft that is the subject of the estimation of the causal relationship among the plurality of feature amounts Ft. As described above, the plurality of feature amounts Ft are feature amounts Ft obtained from at least one time-series data representing one physical quantity possessed by the same measurement target. In this embodiment, the identification image IMf is a rectangular frame image surrounding the plurality of specific nodes NDf. In the causal relationship image IMG shown in FIG. 8, the identification image IMf has a first identification image IMf1 and a second identification image IMf2. The first identification image IMf1 is an image for identifying the plurality of first specific nodes NDf1 as one aggregate. The plurality of first specific nodes NDf1 are nodes ND corresponding to the feature amount Ft that is the subject of the estimation of the causal relationship among the plurality of feature amounts Ft possessed by the time-series data 38 of the pressure data detected from the same measurement target. The first identification image IMf1 is a rectangular frame image surrounding the plurality of first specific nodes NDf1. The second identification image IMf2 is an image for identifying the second specific nodes NDf2 as one collection. The second specific nodes NDf2 are nodes ND corresponding to the feature amount Ft of the causal measurement target among the feature amounts Ft of the time series data 38 of the temperature data detected from the same measurement target. The second identification image IMf2 is a rectangular frame image surrounding the second specific nodes NDf2.

[0051] 9 is a diagram for explaining the causal relationship image IMG. In the causal relationship image IMG, the position of the displayed node ND can be freely changed in the causal relationship image IMG using a cursor or the like. Here, in each of the first identification image IMf1 and the second identification image IMf2, the positions of the multiple first specific nodes NDf1 and multiple second specific nodes NDf2, which are treated as one aggregate, can be changed collectively as one aggregate. For example, when the cursor is moved upward in a state where the frame line or inside of the first identification image IMf1, which is a frame image, is selected using a cursor or the like, the first identification image IMf1 and the multiple first specific nodes NDf1 move upward collectively as shown in FIG. 8 and FIG. 9.

[0052] FIG. 10 is a diagram showing the floating display of the causal relationship image IMG. In the causal relationship image IMG shown in FIG. 8, when a specific node area Rg in which a plurality of specific nodes NDf are displayed is selected with a cursor or the like, the display control unit 28 displays the selected specific node area Rg in a floating manner as shown in FIG. 10. In the causal relationship image IMG shown in FIG. 8, the specific node area Rg has a first specific node area Rg1 in which a plurality of first specific nodes NDf1 are displayed, and a second specific node area Rg2 in which a plurality of second specific nodes NDf2 are displayed. The first specific node area Rg1 is an area inside the frame line and the frame line of the first identification image IMf1. The second specific node area Rg2 is an area inside the frame line and the frame line of the second identification image IMf2. FIG. 10 shows an example in which the first specific node area Rg1 is displayed in a floating manner when the first specific node area Rg1 is selected with a cursor or the like. By displaying an image of the specific node region Rg enlarged in a floating display, the user can easily grasp the specific nodes NDf and extracted edges EGs in the region in which multiple specific nodes NDf are displayed. In particular, when multiple specific nodes NDf are displayed in the specific node region Rg, the floating display makes it easier for the user to grasp the specific nodes NDf and extracted edges EGs.

[0053] 11 is a diagram showing an example of a causal relationship image IMGv of a reference example. The causal relationship image IMGv is an image generated without treating the plurality of first specific nodes NDf1 and the plurality of second specific nodes NDf2 as one aggregate, and does not have an identification image IMf. In the causal relationship image IMGv of the reference example, the first reference value and the second reference value are preset to the same value. In the case of the reference example, the extracted edges EGs between the first specific nodes NDf1 and the extracted edges EGs between the second specific nodes NDf2 are displayed, and the number of extracted edges EGs displayed is greater than the number of extracted edges EGs shown in FIG. 8.

[0054] It is common for there to be a spurious correlation, or so-called multicollinearity, between multiple feature quantities obtained from at least one time-series data representing a single physical quantity of the same measurement target. For example, a large maximum value suggests the presence of a peak, and the absence of a peak leads to a decrease in the average value. Multiple feature quantities obtained from at least one time-series data representing a single physical quantity of the same measurement target are affected by control variables and environmental variables, but are unlikely to have a causal relationship with each other. Therefore, the extracted edges EGs between the nodes ND in the first identification image IMf1 and between the nodes ND in the second identification image IMf2 are of low importance for searching for causal relationships. By displaying the extracted edges EGs between the nodes ND in the first identification image IMf1 and between the nodes ND in the second identification image IMf2, the extracted edges EGs with low importance are displayed in the causal relationship image IMG, which may make it difficult for the user to grasp the causal relationship. Furthermore, since the identification image IMf is not displayed in the causal relationship image IMGv, it is difficult for the user to identify multiple nodes ND generated based on one physical quantity, which makes it difficult for the user to find an edge EG with high importance in the causal relationship image IMGv.

[0055] On the other hand, according to the first embodiment, the first reference value used in the first display condition is higher than the second reference value used in the second display condition. This reduces the number of specific edges EGf connecting specific nodes NDf displayed in the causal relationship image IMG, and allows edges EG with high importance in terms of causal relationships to be displayed. Therefore, the user can easily trace the original cause from the displayed causal relationship image IMG.

[0056] According to the first embodiment, as shown in FIG. 8, the causal relationship image IMG includes an identification image IMf. This allows the user to easily identify, as one aggregate, a plurality of specific nodes NDf corresponding to a plurality of feature quantities Ft obtained from at least one time-series data representing one physical quantity possessed by the same measurement target. Therefore, the user can easily trace the original cause from the causal relationship image IMG by, for example, regarding the edge EG between the plurality of specific nodes NDf displayed as one aggregate by the identification image IMf as being less important than the other edges EG. By being able to easily trace the original cause from the causal relationship image IMG, the user can identify the original cause from the causal relationship image IM in a shorter time.

[0057] According to the first embodiment, the identification image IMf is a frame image as shown in Fig. 8. This allows the user to easily recognize that the multiple specific nodes NDf in the frame image are multiple feature values ​​Ft obtained from at least one time-series data 38 representing one physical quantity of the same measurement target.

[0058] According to the first embodiment, step S20 of generating the causal relationship image IMG includes a process of extracting extracted edges EGs that satisfy a display condition including a first display condition and a second display condition from among a plurality of edges EG connecting between nodes ND, in order to generate the causal relationship image IMG, as shown in steps S23 to S38 in FIG. 7. According to the first embodiment, the display condition is, for example, as shown in FIG. 5, that the absolute value of an index value indicating the strength of the relationship between nodes ND is equal to or greater than a predetermined reference value. As a result, only extracted edges EGs whose absolute values ​​of index values ​​are equal to or greater than the reference value from among the plurality of edges EG are displayed as part of the causal relationship image IMG, so that the number of edges EG displayed can be reduced. This allows a user to easily trace the original cause by referring to the causal relationship image IMG.

[0059] B. Second embodiment: 12 is a diagram showing a causal relationship image IMGa of the second embodiment. The difference between the second embodiment and the first embodiment is that in the second embodiment, a plurality of specific nodes NDf generated based on one physical quantity are imaged as one set node NDg. The other configurations and steps are the same between the first embodiment and the second embodiment, so that the description of the same configurations and steps will be omitted as appropriate.

[0060] The relationship image generating unit 27 in the second embodiment generates a causal relationship image by regarding a plurality of specific nodes NDf as a set node Ndg that is one node ND in step S30 of Fig. 7. In detail, as shown in Fig. 12, the relationship image generating unit 27 generates a causal relationship image IMGa by regarding a plurality of first specific nodes NDf1 as a first set node Ndg1 that is one node and a plurality of second specific nodes NDf2 as a second set node Ndg2 that is one node.

[0061] 13 is a diagram showing a detailed image IMd in the second embodiment. The relationship image generating unit 27 in the second embodiment generates a detailed image IMd including a specific node NDf in the set and an edge EG connected to the specific node NDf for each of the first set node Ndg1 and the second set node Ndg2 as a part of the causal relationship image IMGa. When a set node area Rga in which the set node Ndg is displayed in the causal relationship image IMGa is selected by a user with a cursor or the like, the display control unit 28 causes the display device 50 to display the detailed image IMd instead of the set node Ndg. In detail, when the set node area Rga is selected with a cursor or the like, the display control unit 28 causes the display device 50 to display the detailed image IMd of the selected set node Ndg instead of the causal relationship image IMGa, thereby executing a so-called drill-down display. FIG. 13 shows an example in which the first detailed image IMd1 is displayed instead of the causal relationship image IMGa when the first set node area Rga1 is selected from the first set node area Rga1 and the second set node area Rga2. In this embodiment, the first detailed image IMd1 has a target detailed image IMd1a of the selected set node Ndg and a peripheral image IMd1b showing the connection relationship of the input side node and the output side node to the node ND of the selected set node Ndg. In this embodiment, the set node area Rga is the area inside the frame line and the identification image IMf, which is the identification image IMf. Note that, in other embodiments, the set node area Rga may be an area in which the set node Ndg is displayed.

[0062] According to the second embodiment, the same effects as those of the first embodiment can be obtained in that the configuration and steps are the same. Furthermore, according to the second embodiment, a causal relationship image IMGa is displayed by regarding a plurality of specific nodes NDf as one set node Ndg, so that the user can first grasp the rough causal relationship from the simplified causal relationship image IMGa. Furthermore, according to this embodiment, when a specific node region Rg is selected, a detailed image IMd is displayed, so that the user can easily trace the original cause.

[0063] C. Other ways to reduce visible edges: In the above embodiments, in order to reduce the number of extracted edges EGs between specific nodes NDf included in the causal relationship images IMG and IMGa, the first reference value used for the extracted edges EGs is set to a value higher than the second reference value used for the other edges EGg. However, other methods for reducing the number of extracted edges EGs between specific nodes NDf to be displayed may also be used. Specific embodiments of other methods are described below.

[0064] C-1: First concrete form: The number of extracted edges EGs between specific nodes NDf may be reduced by changing the display condition. For example, the display condition may be a condition that the edge EG to be displayed is another edge EGg different from the specific edge EGf connecting the multiple specific nodes NDf, instead of the first display condition and the second display condition. In other words, the causal relationship image IMG, IMGa does not include the specific edge EGf among the edges EG, but includes only the other edges EGg. The multiple feature quantities Ft obtained from at least one time-series data 38 representing one physical quantity of the same measurement target are unlikely to have a causal relationship with each other. According to this embodiment, by not displaying the specific edge EGf as the causal relationship image IMG, IMGa, the number of displayed specific edges EGf that are less important for searching for causal relationships can be set to zero, while the other edges EGg that are important for searching for causal relationships can be maintained. This allows the user to easily trace the original cause from the displayed causal relationship image IMG, IMGa. Note that instead of displaying all of the other edges EGg in the causal relationship images IMG, IMGa, only the other edges EGg that satisfy a second display condition that defines a second reference value may be displayed, as in the first embodiment.

[0065] C-2: Second specific form: A plurality of nodes ND corresponding to a plurality of feature quantities Ft generated based on one physical quantity may be excluded from the calculation of index values. As a result, an edge EG connecting a plurality of nodes based on one physical quantity is not generated, so that the number of specific edges EGf displayed in the causal relationship images IMG, IMGa that are less important for the search for causal relationships can be reduced, while other edges EGg that are important for the search for causal relationships can be maintained. This allows the user to easily trace the original cause from the displayed causal relationship images IMG, IMGa. Note that instead of displaying all of the other edges EGg in the causal relationship images IMG, IMGa, only other edges EGg that satisfy a second display condition that defines a second reference value may be displayed, as in the first embodiment.

[0066] D. Other embodiments: In each of the above embodiments, the identification image IMf is a frame image as shown in FIG. 8, but is not limited to this as long as the plurality of specific nodes NDf generated based on one physical quantity can be identified as one aggregate. The identification image IMf may be, for example, a color image displayed in the background of an area in which the plurality of specific nodes NDf are located. That is, in the causal relationship images IMG, IMGa, the plurality of specific nodes NDf are displayed superimposed on a color image. Even in this way, the user can easily identify the plurality of specific nodes NDf generated based on one physical quantity as one aggregate. Therefore, the user can easily trace the original cause from the causal relationship images IMG, IMGa.

[0067] E. Other forms: The present disclosure is not limited to the above-mentioned embodiment, and can be realized in various forms without departing from the spirit of the present disclosure. For example, the present disclosure can be realized in the following aspects. The technical features in the above-mentioned embodiments corresponding to the technical features in each aspect described below can be appropriately replaced or combined in order to solve some or all of the problems of the present disclosure, or to achieve some or all of the effects of the present disclosure. Furthermore, if the technical feature is not described as essential in this specification, it can be appropriately deleted.

[0068] (1) According to a first embodiment of the present disclosure, there is provided a method for displaying a causal relationship between a plurality of variables. The plurality of variables include a plurality of feature quantities obtained from at least one time-series data representing a physical quantity possessed by a single measurement target. This method includes (a) a step of generating a causal relationship image having a plurality of nodes representing each of the plurality of variables including the plurality of feature quantities, extraction edges connecting the nodes and satisfying a predetermined display condition, and an identification image for identifying a plurality of specific nodes among the plurality of nodes corresponding to the plurality of feature quantities as one collection; and (b) a step of displaying the generated causal relationship image on a display device. According to this embodiment, a user can easily identify a plurality of specific nodes corresponding to a plurality of feature quantities obtained from at least one time-series data representing a physical quantity possessed by a single measurement target as one collection. Thus, the user can easily trace the original cause from the causal relationship image.

[0069] (2) In the above embodiment, the identification image may be a frame image surrounding the specific nodes. With this embodiment, a user can easily recognize that the specific nodes within the frame image are multiple feature quantities obtained from at least one time-series data representing one physical quantity of a single measurement target.

[0070] (3) In the above embodiment, the step (a) may include a step of extracting the extracted edges that satisfy the display condition from among the plurality of edges connecting the nodes in order to generate the causal relationship image, and the display condition may be a condition that an absolute value of an index value indicating the strength of the relationship between the nodes is equal to or greater than a predetermined reference value. According to this embodiment, only the extracted edges whose absolute values ​​of index values ​​are equal to or greater than a reference value are displayed as part of the causal relationship image, so that the number of edges to be displayed can be reduced. This allows a user to easily trace the original cause by referring to the causal relationship image.

[0071] (4) In the above embodiment, the step (a) may include a step of extracting the extracted edges that satisfy the display condition from among the plurality of edges connecting the nodes in order to generate the causal relationship image, and the display condition may include a condition that the edges that connect the nodes are different from the specific edges connecting the plurality of specific nodes. A spurious correlation, so-called multicollinearity, is generally present between a plurality of feature quantities obtained from at least one time-series data representing a single physical quantity possessed by the same measurement object. A plurality of feature quantities obtained from at least one time-series data representing a single physical quantity possessed by the same measurement object is unlikely to have a causal relationship with each other. According to this embodiment, by not displaying the specific edges as the causal relationship image, it is possible to reduce the number of edges to be displayed while maintaining the display of edges that are highly important in terms of causal relationships. This allows a user to easily trace the original cause from the displayed causal relationship image.

[0072] (5) In the above embodiment, the step (a) includes a step of extracting the extracted edges that satisfy the display condition from among the plurality of edges connecting the nodes in order to generate the causal relationship image, and the display condition includes a first display condition used for a specific edge connecting the plurality of specific nodes among the plurality of edges connecting the nodes, and a second display condition used for the edge different from the specific edge, and the first display condition may be a condition that an absolute value of an index value indicating the strength of the relationship between the nodes is equal to or greater than a first reference value that is determined in advance, and the second display condition may be a condition that the absolute value of the index value is equal to or greater than a second reference value that is lower than the first reference value. According to this embodiment, the first reference value used in the first display condition is higher than the second reference value used in the second display condition. This reduces the number of specific edges connecting the specific nodes displayed in the causal relationship image, and maintains the display of edges that are important in terms of causal relationships. Therefore, a user can easily trace the original cause from the displayed causal relationship image.

[0073] (6) In the above embodiment, the index value may be a value using any one of the following indices: mutual information, Pearson's product moment correlation, Hilbert-Schmidt independence criterion, partial correlation coefficient, correlation ratio, association coefficient, rank correlation, and entropy. According to this embodiment, the index value can be easily calculated using the above values.

[0074] (7) In the above embodiment, the step (b) may include a step of floating-displaying a specific node region, which is a region in which the multiple specific nodes are displayed in the causal relationship image, when the specific node region is selected. According to this embodiment, the floating display makes it easier for a user to grasp the specific nodes and edges in the region in which the multiple specific nodes are displayed.

[0075] (8) In the above embodiment, the step (a) may generate the causal relationship image by regarding the specific nodes as a set node that is one of the nodes, and the method may further include a step of (c) generating a detailed image including the specific nodes and the edges connected to the specific nodes, and a step of (d) displaying the detailed image on the display device when a set node area in which the set node is displayed in the causal relationship image is selected. According to this embodiment, the causal relationship image is displayed by regarding the specific nodes as one set node, so that the user can first grasp the general causal relationship from the simplified causal relationship image. Also, according to this embodiment, when a specific node area is selected, a detailed image is displayed, so that the user can trace the original cause.

[0076] (9) According to a second aspect of the present disclosure, there is provided a computer program for causing a computer to execute display of a causal relationship between a plurality of variables on a display device. The plurality of variables include a plurality of feature quantities obtained from at least one time-series data representing a physical quantity obtained from the same measurement target. This computer program includes (a) a function for generating a causal relationship image having a plurality of nodes representing each of the plurality of variables including the plurality of feature quantities, extraction edges which are edges connecting the nodes and satisfy a predetermined display condition, and an identification image for identifying a plurality of specific nodes among the plurality of nodes corresponding to the plurality of feature quantities as one collection, and (b) a function for displaying the generated causal relationship image on a display device. According to this aspect, a user can easily identify a plurality of specific nodes corresponding to a plurality of feature quantities obtained from at least one time-series data representing a physical quantity possessed by the same measurement target as one collection. Thus, a user can easily trace the original cause from the causal relationship image.

[0077] The present disclosure may be realized in various forms other than those described above, for example, as a computer program or a non-transitory storage medium having a computer program recorded thereon. [Explanation of symbols]

[0078] Bx...checkbox image, EG...edge, EGf...specific edge, EGg...other edge, EGs...extracted edge, Ft, Ft_A0~Ft_A5, Ft_B0~Ft_B5...feature, Ft_av...average value, Ft_in...integral value, Ft_max...maximum value, Ft_min...minimum value, IM1...first setting image, IM2...second setting image, IM2a...variable selection image, IM2b...covariate selection image, IM3...third setting image, IM3a...first reference value setting image, IM3b...second reference value setting image, IMG, IMGa, IMGv...causal relationship image, IMGr1, IMGr2...individual image, IMI...individual setting image, IMIa...first individual setting image, IMIb...second individual setting image, IMSa, IMSb, IMSc...instruction input image, IMd...detailed image, IMd1...first detailed image, IMd1a...target detailed image, IMd1b...surrounding image, Mf...identification image, IMf1...first identification image, IMf2...second identification image, IMra...first bar image, IMrb...first value display image, IMt...order setting image, IMta~IMtd...images, IMva...second bar image, IMvb...second value display image, ND...node, NDf...specific node, NDf1...first specific node, NDf2...second specific node, Ndg...set node, Ndg1...first set node, Ndg2...second set node, Rg...specific node area, Rg1...first specific node area, Rg2...second specific node area, Rga...set node area, Rga1...first set node area, Rga2...second set node area

Claims

1. 1. A method for displaying causal relationships of multiple variables, comprising: the plurality of variables include a plurality of feature quantities obtained from at least one piece of time-series data representing a physical quantity of a single measurement target, (a) generating a causal relationship image having a plurality of nodes each representing a plurality of variables including the plurality of feature quantities, extraction edges connecting the nodes and satisfying a predetermined display condition, and an identification image for identifying a plurality of specific nodes corresponding to the plurality of feature quantities among the plurality of nodes as one collection; (b) displaying the generated causal relationship image on a display device.

2. 2. The method of claim 1 , The method, wherein the identifying image is a frame image surrounding the plurality of identified nodes.

3. 2. The method of claim 1 , The step (a) extracting the extracted edges that satisfy the display condition from among the plurality of edges connecting the nodes in order to generate the causal relationship image; The display condition is that an absolute value of an index value indicating a strength of a relationship between the nodes is equal to or greater than a predetermined reference value.

4. 2. The method of claim 1 , The step (a) extracting the extracted edges that satisfy the display condition from among the plurality of edges connecting the nodes in order to generate the causal relationship image; The method according to the present invention, wherein the display condition includes a condition that the edge, among the plurality of edges connecting the nodes, is different from a specific edge connecting the plurality of specific nodes.

5. 2. The method of claim 1 , The step (a) extracting the extracted edges that satisfy the display condition from among the plurality of edges connecting the nodes in order to generate the causal relationship image; The display condition is, among the plurality of edges connecting the nodes, a first display condition used for a specific edge connecting the plurality of specific nodes; a second display condition used for the edge different from the specific edge, the first display condition is a condition that an absolute value of an index value indicating a strength of a relationship between the nodes is equal to or greater than a predetermined first reference value, The method, wherein the second display condition is that the absolute value of the index value is equal to or greater than a second predetermined reference value that is lower than the first reference value.

6. 6. The method of claim 3 or claim 5, The index value is a value using any one of mutual information, Pearson's product-moment correlation, Hilbert-Schmidt independence criterion, partial correlation coefficient, correlation ratio, association coefficient, rank correlation, and entropy.

7. 2. The method of claim 1 , The method, wherein step (b) includes a step of, when a specific node area in which the plurality of specific nodes are displayed in the causal relationship image is selected, displaying the specific node area in a floating manner.

8. 2. The method of claim 1 , The step (a) generates the causal relationship image by regarding the plurality of specific nodes as a set node that is one of the nodes; The method further comprises: (c) generating a detailed image including the plurality of identified nodes and the edges connected to the identified nodes; (d) causing the display device to display the detailed image when a set node region in which the set node is displayed in the causal relationship image is selected.

9. A computer program for causing a computer to display on a display device a causal relationship between a plurality of variables, comprising: the plurality of variables include a plurality of feature quantities obtained from at least one piece of time-series data representing a physical quantity acquired from a single measurement target; (a) a function of generating a causal relationship image having a plurality of nodes representing each of the plurality of variables including the plurality of feature quantities, extracted edges which are edges connecting the nodes and satisfy a predetermined display condition, and an identification image for identifying a plurality of specific nodes corresponding to the plurality of feature quantities among the plurality of nodes as one collection; (b) a function for displaying the generated causal relationship image on a display device.

Images