Information processing apparatus, information processing system, information processing method, and program

Abstract

To specify a cause affecting fluctuation of data.SOLUTION: The information processing apparatus includes a processing unit. The processing unit calculates a plurality of external noise estimation values corresponding to the plurality of variables for each of one or more pieces of result data based on the one or more pieces of result data including a plurality of result values respectively corresponding to the plurality of variables and a structural causal model representing a causal relationship of the plurality of variables. Each of the plurality of external noise estimation values represents an estimation value of an influence due to an external noise different from influences from the plurality of variables on a corresponding variable among the plurality of variables. The processing unit generates, based on the structural causal model and the plurality of external noise estimation values for each of the one or more pieces of result data, for each of the one or more pieces of result data, for each combination of two variables included in the plurality of variables, a degree of contribution representing a magnitude of an influence of an external noise given to a source variable, which is one of the two variables, on a target variable, which is the other.SELECTED DRAWING: Figure 4

Description

[Technical Field]

[0001] The present invention relates to an information processing device, an information processing system, an information processing method, and a program. [Background technology]

[0002] Techniques have been proposed that use statistics and machine learning to unravel the complex structures hidden in data, such as causal discovery and causal inference, to analyze causal relationships in data and improve prediction and decision-making methods.

[0003] For example, in manufacturing systems, causal discovery and inference are used to identify factors that affect product quality and predict the impact of process changes on product quality based on raw material data, process data, quality inspection data, and maintenance data, etc. In information systems, causal relationships between system components are identified to perform fault analysis and performance improvement. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] JP 2009-54843 A [Patent Document 2] Patent No. 5108116 [Patent Document 3] Patent No. 6937345 [Patent Document 4] Patent No. 7173395 [Patent Document 5] Patent No. 7251687 [Patent Document 6] Patent No. 5825599 [Patent Document 7] Patent Publication No. 2023-173391 [Patent Document 8] International Publication No. 2022 / 259446 [Non-patent literature]

[0005] [Non-Patent Document 1] Peters, J., Janzing, D., & Scholkopf, B., “Elements of causal inference: foundations and learning algorithms”, 2017, The MIT Press, pp.81-155 [Non-patent document 2] Shimizu, S., Hoyer, PO, Hyvarinen, A., Kerminen, A., & Jordan, M., “A linear non-Gaussian acyclic model for causal discovery”, published 2006, Journal of Machine Learning Research, 7(10) [Non-patent document 3] Shimizu, S., Inazumi, T., Sogawa, Y., Hyvarinen, A., Kawahara, Y., Washio, T., & Hoyer,PO, “DirectLiNGAM: A direct method for learning a linear non-Gaussian structural equation model”, published 2011, Journal of Machine Learning Research-JMLR, 12(Apr), pp.1225-1248 [Non-patent document 4] Hyvarinen, A., & Smith, SM, “Pairwise likelihood ratios for estimation of non-Gaussian structural equation models”, 2013, The Journal of Machine Learning Research, 14(1), pp.111-152 [Non-Patent Document 5] Hoyer, P., Janzing, D., Mooij, JM, Peters, J., & Scholkopf, B., “Nonlinear causal discovery with additive noise models”, 2008, Advances in neural information processing systems, 21 [Non-patent document 6] Peters, J., Mooij, JM, Janzing, D., & Scholkopf, B., “Causal Discovery with Continuous Additive Noise Models”, published 2014, Journal of Machine Learning Research, 15, pp.2009-2053 [Non-Patent Document 7] Peters, J., & Buhlmann, P., “Identifiability of Gaussian structural equation models with equal error variances”, 2014, Biometrika, 101(1), pp.219-228 [Non-patent document 8] Vukovic, M., & Thalmann, S., “Causal discovery in manufacturing: A structured literature review”, published in 2022, Journal of Manufacturing and Materials Processing, 6(1), 10 [Non-Patent Document 9] Marazopoulou, K., Ghosh, R., Lade, P., & Jensen, D., “Causal discovery for manufacturing domains”, published 2016, arXiv preprint arXiv:1605.04056 [Non-Patent Document 10] Budhathoki, K., Minorics, L., Blobaum, P., & Janzing, D., “Causal structure-based root cause analysis of outliers”, June 2022, In International Conference on Machine Learning, pp.2357-2369, PMLR [Non-Patent Document 11] Strobl, EV, & Lasko, TA “Identifying patient-specific root causes of disease”, published August 2022, In Proceedings of the 13th ACM International Conference on Bioinformatics, Computational Biology and Health Informatics, pp.1-10 [Non-Patent Document 12] Strobl, E., & Lasko, TA, “Sample-specific root causal inference with latent variables”, August 2023, In Conference on Causal Learning and Reasoning, pp.895-915, PMLR Summary of the Invention [Problem to be solved by the invention]

[0006] The problem that the present invention aims to solve is to identify the causes that affect the fluctuations in data. [Means for solving the problem]

[0007] An information processing device according to an embodiment includes a processing unit. The processing unit calculates, for each of the one or more pieces of performance data, a plurality of external noise estimates corresponding to the plurality of variables, based on one or more pieces of performance data including a plurality of performance values ​​each corresponding to the plurality of variables and a structural causal model representing a causal relationship between the plurality of variables. Each of the plurality of external noise estimates represents an estimate of an influence of external noise different from the influence of the plurality of variables on a corresponding variable among the plurality of variables. The processing unit generates, for each of the one or more pieces of performance data, a contribution representing an influence of the external noise applied to a source variable, which is one of the two variables, on a target variable, which is the other of the two variables, based on the structural causal model and the plurality of external noise estimates for each of the one or more pieces of performance data. [Brief explanation of the drawings]

[0008] [Figure 1] FIG. 1 is a configuration diagram of an information processing system according to an embodiment. [Figure 2] Diagram showing the structural causal model. [Figure 3] Diagram showing a causal graph. [Figure 4] FIG. [Figure 5] FIG. [Figure 6] FIG. 10 is a diagram showing an example of an adjacency matrix. [Figure 7] FIG. 10 is a diagram showing an extrinsic noise matrix. [Figure 8] Diagram showing a third-order tensor. [Figure 9] FIG. 10 is a diagram showing the results of a comparison between the analysis device and a conventional regression model. [Figure 10] FIG. 2 is a diagram showing an example of the configuration of an output unit. [Figure 11] FIG. 10 is a diagram showing a first example of cause information. [Figure 12] FIG. 10 is a diagram showing a second example of cause information. [Figure 13] FIG. 10 is a diagram showing a third example of cause information. [Figure 14] FIG. 10 is a diagram showing a fourth example of cause information. [Figure 15] FIG. 10 is a diagram showing a fifth example of cause information. [Figure 16] FIG. 10 is a diagram showing a sixth example of cause information. [Figure 17] FIG. 13 is a diagram showing a seventh example of cause information. [Figure 18] FIG. 13 is a diagram showing an eighth example of cause information. [Figure 19] FIG. 13 is a diagram showing a ninth example of cause information. [Figure 20] FIG. 10 is a diagram showing a first display example of a causal graph and cause information. [Figure 21] FIG. 10 is a diagram showing a second display example of a causal graph and cause information. [Figure 22] FIG. 10 is a diagram showing a third display example of a causal graph and cause information. [Figure 23] FIG. 10 is a diagram showing a modified example of the information processing system according to the embodiment. [Figure 24] FIG. 2 is a diagram illustrating the hardware configuration of the analysis device. DETAILED DESCRIPTION OF THE INVENTION

[0009] 1 is a diagram showing the configuration of an information processing system 10 according to the embodiment. The information processing system 10 according to the embodiment includes a target system 20 and an analysis device 30.

[0010] The target system 20 is, for example, a manufacturing system that manufactures products. The target system 20 may also be a data processing system that executes computer processes, or an information processing system that provides information processing services using information processing. The target system 20 is not limited to such systems, and may be any system that handles data.

[0011] The analysis device 30 is an information processing device that executes information processing. The analysis device 30 acquires one or more pieces of performance data including a plurality of performance values ​​each corresponding to a plurality of variables from the target system 20. A structural causal model (SCM) is set in advance in the analysis device 30, for example, when the target system 20 starts operating, is initialized, or is shipped from the factory.

[0012] The analysis device 30 generates cause information indicating the cause of a change in at least one of the multiple variables based on the acquired one or more pieces of performance data and a predetermined structural causal model. The analysis device 30 then outputs the cause information, for example, by displaying it on a display device. For example, in order to manage the state of the target system 20, the analysis device 30 periodically acquires one or more pieces of performance data including the performance values ​​of the multiple variables as a routine task and outputs the cause information.

[0013] Each of the multiple variables represents a value sampled in the target system 20. For example, if the target system 20 is a manufacturing system, each of the multiple variables represents raw material data such as the amount and quality of raw materials, process data such as the operating time of equipment during manufacturing and sensor data detected by a sensor on the environment of the manufacturing equipment, quality inspection data indicating the quality of manufactured products, and maintenance data detected during maintenance.

[0014] Each of the one or more pieces of performance data is a vector including a plurality of performance values. In this embodiment, each of the one or more pieces of performance data is a vector including d variables (X1, X2, ..., X d ) is a d-dimensional vector containing d actual values ​​that correspond one-to-one to the

[0015] Furthermore, each of the one or more pieces of performance data includes a plurality of performance values ​​sampled from the target system 20 under conditions such as different times. For example, the first performance data and the second performance data among the one or more pieces of performance data are values ​​sampled from the target system 20 at different times. However, the multiple performance values ​​included in one piece of performance data are values ​​sampled under the same conditions such as the same time.

[0016] In this embodiment, the analysis device 30 acquires n samples of performance data (n is an integer equal to or greater than 1). In this embodiment, each of the one or more performance data is assigned an index that identifies the conditions, such as the time at which it was sampled.

[0017] A structural causal model is information that represents the causal relationships among multiple variables. That is, a structural causal model is information that, for each pair of two variables included in the multiple variables, represents whether one variable in the pair affects the other variable and the magnitude of the influence.

[0018] In this embodiment, the structural causal model is a linear model in which the magnitude of influence is expressed by real numbers, and is expressed using an adjacency matrix B. For each combination of two variables included in the plurality of variables, the adjacency matrix B represents the magnitude of influence from one variable to the other. In this embodiment, the number of variables is d, and the adjacency matrix B is expressed as a square matrix with d rows and d columns.

[0019] Each element included in adjacency matrix B contains a real value representing the magnitude of influence from a variable identified by a row (one variable) to a variable identified by a column (the other variable). The magnitude of influence from one variable to another variable may be positive, negative, or 0. 0 indicates that one variable has no influence on the other variable. Note that adjacency matrix B contains the magnitude of influence of pairs of variables where one variable and the other variable are the same. The magnitude of influence of pairs where one variable and the other variable are the same is included in the diagonal elements of adjacency matrix B and is 0. Note that the rows and columns of adjacency matrix B may be reversed from the example of this embodiment.

[0020] Here, the index of the variable (X d Simple regression analysis and correlation analysis are known methods for analyzing the causes of changes in .

[0021] The variable vector X is expressed by equation (1). d ) represent real numbers. The superscript T represents a transposed matrix.

number

[0022] For example, the variable (X j ) as explanatory variables, and variables with index d (X d The simple regression model with the objective variable is expressed as equation (2).

number

[0023] In equation (2), ε j is the variable with index j (X j ) is the noise given to β 0j and β 1j are the coefficients of the explanatory variables and are the parameters of the simple regression model.

[0024] β 0j and β 1j is estimated by the least squares method using the actual values ​​of d variables. The coefficient of determination of such a simple regression model is equal to the square of the correlation coefficient. Therefore, in simple regression analysis and correlation analysis, it is important to select explanatory variables (X j ) is the response variable (X d ) is analyzed to have a high correlation with

[0025] Also, for example, if the index of a variable vector X containing d variables is d, thend Another traditional method for analyzing the causes of changes in variables (X1, X2, …, X) is multiple regression analysis. d-1 ,) as explanatory variables, and the variables with index d (X d The multiple regression model with the objective variable is expressed as in equation (3).

number

[0026] In equation (3), ε d is the variable with index d (X d ) are the noises given to β0, β1, …β d-1 are the coefficients of the explanatory variables and are the parameters of the multiple regression model.

[0027] β0, β1, … β d-1 is estimated by the least squares method using actual values ​​of d variables. When there are many explanatory variables, the multiple regression model may be expressed using a portion of the d variables selected using domain knowledge. Furthermore, the multiple regression model may be expressed by selecting a portion of the d variables, reducing the dimensionality, or reducing and reducing the dimensionality using a data-driven method such as Ridge, Lasso, PCA, or PLS.

[0028] The analysis results using such a simple regression model are as follows: j ) is the actual value of x j But the variable with index j (X j ) is the average value of X j_ave From (x j -X j_ave ), ignoring the influence of other variables, the target variable (X d ) is the actual value of x d But +β 1j (x j -X j_ave ) tends to change by the same amount. Also, the results of the analysis using the multiple regression model show that "the variable with index j (X j) is the actual value of x j But the variable with index j (X j ) is the average value of X j_ave From (x j -X j_ave ), when other variables are fixed, the target variable (X d ) is the actual value of x d But +β 1j (x j -X j_ave ) tends to change by a factor of 1.

[0029] However, the analysis results using the simple regression model and the multiple regression model only reflect statistical correlations, and do not guarantee causal relationships between variables. For this reason, the analysis results using the simple regression model and the multiple regression model are not necessarily based on the objective variable (X d ) is difficult to determine the causal cause of the change. Furthermore, the analysis results using a simple regression model do not take into account the influence of other variables. Also, the analysis results using a multiple regression model analyze partial correlations by fixing other variables, so X j It only reflects the influence of variables that directly affect the

[0030] In contrast, the analysis device 30 according to this embodiment can output cause information indicating indirect causes in addition to direct causes of changes in variables by using a structural causal model, as described below.

[0031] Figure 2 is a diagram of the structural causal model.

[0032] In this embodiment, the structural causal model is expressed as in equation (4).

number

[0033] k, j are the d variables (X1,X2,…,X d ) is an index that identifies one of the variables in X. kis a function of d variables (X1,X2,…,X d ) with index k (X k ) value. B jk is a real value. B jk is a function of d variables (X1,X2,…,X d ), the variable with index j (X j ) and the variable with index k (X k ) and the variable with index j (X j ) to the variable with index k (X k ) and P a (k) is the variable with index k (X k ) represents the set of indices of parent variables that directly influence the

[0034] E k is the variable with index k (X k ) represents the exogenous noise added to the model. Exogenous noise is noise caused by external factors other than the influence of multiple variables.

[0035] When expressed as a matrix, the structural causal model is expressed as in equation (5).

number

[0036] E is the sum of d external noises (E1, E2, ..., E d ) is a vector containing

number

[0037] X is a set of d variables (X1, X2, …, X d ) is a vector containing

[0038] B T is the transposed matrix of the adjacency matrix B. As shown in equation (7), the adjacency matrix B has d×d elements whose values ​​are real numbers.

number

[0039] In addition, B T is sometimes called an adjacency matrix, but in this embodiment, B is used as the adjacency matrix.

[0040] A structural causal model can also be expressed as multiple equations, as shown on the left side of Figure 2. A structural causal model is also called a structural equation model (SEM).

[0041] FIG. 3 is a diagram illustrating a causal graph representing a structural causal model.

[0042] A structural causal model is represented by a directed causal graph. A causal graph is a set of d variables (X1, ..., X d ) in a one-to-one correspondence.

[0043] B is an element of the adjacency matrix B jk is the variable with index j in the causal graph (X j ) from the node corresponding to the variable with index k (X k ) represents the value corresponding to the directed edge to the node corresponding to

[0044] The causal graph is B jk is nonzero, the variable with index j (X j ) from the node corresponding to the variable with index k (X k ) and the causal graph contains a directed edge to the node corresponding to B jk If is zero, then the variable with index j (X j ) from the node corresponding to the variable with index k (X k ) does not contain a directed edge to the node corresponding to

[0045] Also, E k is the variable with index k (X k ) represents the exogenous noise affecting the corresponding node.

[0046] The analysis device 30 according to this embodiment uses the structural causal model described above to output cause information indicating the direct and indirect causes of a change in a variable.

[0047] Fig. 4 is a diagram showing the configuration of the analysis device 30. In the description of Fig. 4, Figs. 5 to 8 will be referred to. Fig. 5 is a diagram showing an actual data matrix X'. Fig. 6 is a diagram showing an example of an adjacency matrix B. Fig. 7 is a diagram showing an extrinsic noise matrix E'. Fig. 8 is a diagram showing a contribution tensor C.

[0048] The analysis device 30 includes an actual data acquisition unit 32, an actual data storage unit 34, a model acquisition unit 36, a model storage unit 38, an external noise estimation unit 40, a contribution decomposition unit 42, a decomposition result storage unit 44, and an output unit 46.

[0049] The performance data acquiring unit 32 acquires one or more performance data from the target system 20. In this embodiment, the performance data acquiring unit 32 acquires performance data including d performance values ​​corresponding to d variables, for n samples.

[0050] For example, when an intercept exists in the structural causal model, the performance data acquiring unit 32 may perform a predetermined process such as centering or standardization on the multiple performance values ​​included in each of the one or more performance data. Furthermore, the performance data acquiring unit 32 may add a variable with a fixed value.

[0051] The performance data storage unit 34 stores one or more pieces of acquired performance data. In this embodiment, the performance data storage unit 34 stores a performance data matrix X' including n rows corresponding to n samples and d columns corresponding to d variables, as shown in Fig. 5. The elements included in the i-th row and j-th column of the performance data matrix X' are x, which are performance values ​​corresponding to the variable with index j among the d variables included in the sample with index i among the n samples. ijIn the performance data matrix X', i is an integer greater than or equal to 1 and less than or equal to n, and j is an integer greater than or equal to 1 and less than or equal to d.

[0052] The model acquiring unit 36 ​​acquires a structural causal model. In this embodiment, the model acquiring unit 36 ​​acquires an adjacency matrix B. The adjacency matrix B is generated in advance by a device other than the analysis device 30 or by a user. Note that the adjacency matrix B may be generated by the model acquiring unit 36 ​​based on one or more pieces of performance data.

[0053] The adjacency matrix B is estimated based on one or more pieces of performance data using, for example, a causal search algorithm. For example, the adjacency matrix B may be estimated using an algorithm such as multiple regression or Adaptive Lasso after specifying a causal order using domain knowledge.

[0054] Furthermore, for example, the adjacency matrix B may be estimated using known causal structure and covariance structure analysis based on one or more pieces of performance data. For example, the adjacency matrix B may be estimated using information regarding the presence or absence of causality between variables and a causality search algorithm based on one or more pieces of performance data. For example, the adjacency matrix B may be estimated using a causality search algorithm under assumptions such as non-Gaussianity, non-linearity, or homoscedasticity. Non-Gaussianity is described in Non-Patent Documents 2, 3, and 4. Non-linearity is described in Non-Patent Documents 5 and 6. Homoscedasticity is described in Non-Patent Document 7.

[0055] The one or more pieces of performance data used to estimate the adjacency matrix B may be data acquired by the analysis device 30, or may be data different from the data acquired by the analysis device 30, for example, data from a normal period in the past.

[0056] When d=5, the adjacency matrix B is expressed as shown in Figure 6. Each element included in the adjacency matrix B represents the magnitude of the influence from one variable specified by the row to another variable specified by the column.

[0057] The model storage unit 38 stores the structural causal model acquired by the model acquisition unit 36. In this embodiment, the model storage unit 38 stores the adjacency matrix B.

[0058] The external noise estimation unit 40 calculates a plurality of external noise estimation values ​​corresponding to a plurality of variables for each of the one or more pieces of actual data based on the one or more pieces of actual data stored in the actual data storage unit 34 and the structural causal model acquired by the model acquisition unit 36. The plurality of external noise estimation values ​​correspond one-to-one to the plurality of variables. Each of the plurality of external noise estimation values ​​represents an estimate of the influence of the external noise on a corresponding variable among the plurality of variables.

[0059] In this embodiment, the external noise estimation unit 40 generates an external noise matrix E'. The external noise matrix E' includes a plurality of external noise estimated values ​​for each of one or more pieces of performance data.

[0060] For example, the extrinsic noise matrix E' includes n rows corresponding to n samples and d columns corresponding to d variables, as shown in Fig. 7. The element in the i-th row and j-th column of the extrinsic noise matrix E' is an extrinsic noise estimate e that estimates the extrinsic noise given to the variable with index j among the d variables included in the sample with index i among the n samples. ij Includes: e ij is expressed by a real number.

[0061] The external noise estimation unit 40 calculates the external noise matrix E' using equation (8).

number

[0062] Here, I represents a unit matrix. That is, the external noise estimation unit 40 calculates the external noise matrix E' by multiplying the result data matrix X' by a matrix (IB) obtained by subtracting the adjacent matrix B from the unit matrix I.

[0063] Based on the structural causal model and multiple exogenous noise estimates for each of the one or more actual data, the contribution decomposition unit 42 generates, for each combination of two variables included in the multiple variables, a contribution that represents the magnitude of the influence that exogenous noise given to a source variable, one of the two variables, has on a target variable, the other of the two variables, for each of the one or more actual data.

[0064] The contribution will be further explained below.

[0065] By rearranging equation (8), the performance data matrix X' is expressed as in equation (9).

number

[0066] (IB) -1 When is defined as the coefficient matrix A, the performance data matrix X' is expressed as shown in equation (10).

number

[0067] If the causal graph is an acyclic graph, by appropriately rearranging the order of the variables, the adjacency matrix B becomes a strictly upper triangular matrix with zero diagonal elements. Therefore, (IB) has an inverse matrix. That is, the coefficient matrix A = (IB), which is the inverse matrix of (IB). -1 is an upper triangular matrix with diagonal elements of 1. The order of multiple variables when arranged in this way is called the causal order. A causal graph representing a structural causal model may contain a directed edge from a node corresponding to any first-order variable to a node corresponding to any second-order variable whose causal order is greater than first. However, the causal graph does not contain a directed edge from a node corresponding to a second-order variable to a node corresponding to the first-order variable.

[0068] Also, if the causal graph is not acyclic, (IB) does not necessarily have an inverse matrix. Therefore, the coefficient matrix A is (IB) -1 Instead, we use the generalized inverse matrix (IB) + may be.

[0069] The element in row i and column k (k is an integer between 1 and d) in X' = E'A is x ik x ik is expressed as in equation (11).

number

[0070] x represented by equation (11) ik represents the actual value of the variable with index k out of d variables that is included in the actual data of the sample with index i out of n samples.

[0071] The right side of equation (11) is a linear sum of multiple terms. ik is expressed as a linear summation of multiple terms.

[0072] The linear sum equation on the right side of equation (11) is (k-1) A jk e ij and one e ik This includes the terms

[0073] e ik represents the value of the element in row i and column k in the extrinsic noise matrix E'. ik is an exogenous noise estimate representing the exogenous noise given to the variable with index k in the historical data of the sample with index i.

[0074] e ij represents the value of the element in the i-th row and j-th column of the extrinsic noise matrix E'. ij is an exogenous noise estimate representing the exogenous noise given to the variable with index j in the historical data of the sample with index i.

[0075] A jk is the coefficient matrix A=(IB) -1 represents the value of the element in row j and column k in

[0076] Here, in equation (11), the variable with index j is located upstream in the causal graph from the variable with index k. In other words, the variable with index j has a lower causal order than the variable with index k. Therefore, e ij is the actual value of the variable with index k in the sample with index i, x ik The external noise estimate represents the external noise given to a variable that may directly affect the variable, or the external noise estimate represents the external noise given to a variable that may indirectly affect the variable.

[0077] In other words, A in the linear sum equation on the right side of equation (11) jk e ij is the actual value of the variable with index k in the sample with index i, x ik It represents the influence of exogenous noise on variables that may directly or indirectly affect the

[0078] Note that when the adjacency matrix B is a strict upper triangular matrix with diagonal elements set to zero, the coefficient matrix A becomes an upper triangular matrix with diagonal elements set to 1 and elements below the diagonal elements set to 0. Therefore, in this case, equation (11) may be expressed as equation (12).

number

[0079] The linear sum expression on the right side of equation (12) is the multiple exogenous noise estimates e for the actual data of sample with index i. ij and the corresponding coefficients in the coefficient matrix A. jkThat is, the linear sum equation on the right side of equation (12) is an equation that performs a product-sum operation on the row corresponding to the actual data of the sample with index i in the extrinsic noise matrix E' and the column corresponding to the variable with index k in the coefficient matrix A. However, since the coefficient matrix A is an upper triangular matrix with 0 below the diagonal elements, A in the linear sum equation on the right side of equation (12) jk e ij is 0 if j is greater than k.

[0080] The contribution decomposition unit 42 calculates A jk e ij is generated as a contribution for each combination of two variables included in the plurality of variables for each of one or more performance data. That is, the contribution decomposition unit 42 generates A jk e ij Let x be the actual value of the variable with index k contained in the sample with index i. ik for the variable with index k contained in the sample with index i. ik The contribution decomposition unit 42 generates the contribution as the degree of contribution of external noise given to a source variable, which is one of two variables, on a target variable, which is the other of the two variables. Note that the contribution decomposition unit 42 also generates the contribution for a combination of two variables where the source variable and the target variable are the same variable.

[0081] For example, when a first variable among a plurality of variables in a first piece of one or more pieces of actual data is set as a target variable and a second variable among the plurality of variables is set as a source variable, the contribution to the combination of the two variables is the value of a term among a plurality of terms included in a linear sum formula, the term including an extrinsic noise estimated value representing the extrinsic noise given to the second variable. In this case, the linear sum formula is a formula that performs a product-sum operation on the row corresponding to the first actual data in the extrinsic noise matrix E' and the column corresponding to the first variable in the coefficient matrix A. Furthermore, the coefficient matrix A is the inverse matrix (IB) of the matrix (IB) obtained by subtracting the adjacent matrix B from the unit matrix I.-1 or generalized inverse (IB) + is.

[0082] For example, for the actual data of the sample with index i (first actual data), if the variable with index k (first variable) is the target variable and the variable with index j (second variable) is the source variable, the contribution is A jk e ij For the actual data of the sample with index i (first actual data), if the target variable and source variable are both variables with index k (first variable), the contribution is e ik This becomes:

[0083] In this embodiment, the contribution decomposition unit 42 generates a contribution tensor C, which is a third-order tensor. The contribution tensor C includes an i-component, a j-component, and a k-component.

[0084] The i component corresponds to each of one or more pieces of performance data. In this embodiment, the i component corresponds to each of the performance data of n samples.

[0085] The j component and the k component correspond to each of a plurality of variables. In this embodiment, the j component and the k component correspond to each of the d variables.

[0086] In this embodiment, the contribution decomposition unit 42 decomposes C into i, j, and k components of the contribution tensor C. ijk is calculated using equation (13).

number

[0087] That is, as shown in FIG. 8, each of the multiple elements included in the contribution tensor C represents the contribution of one or more pieces of actual data identified by the i component to a variable identified by the j component among the multiple variables as a source variable and a variable identified by the k component among the multiple variables as a target variable.

[0088] For the multiple elements contained in such a contribution tensor C, equation (14) holds.

number

[0089] That is, each of the multiple elements included in the contribution tensor C represents a term included in the linear sum equation for each of one or more pieces of performance data. Therefore, by calculating such a contribution tensor C, the contribution decomposition unit 42 can calculate the results of a comprehensive analysis of the presence of other variables that cause a change in at least one variable among the multiple variables.

[0090] The contribution decomposition unit 42 may include values ​​obtained by converting the numerical value shown in formula (13) according to a predetermined rule in the elements of the contribution tensor C. For example, the contribution decomposition unit 42 may include a sign indicating whether the numerical value shown in formula (13) is positive or negative, an absolute value, a value indicating whether the numerical value is included in a range from a predetermined upper limit value to a predetermined lower limit value, or a value obtained by leveling the numerical value shown in formula (13) into a predetermined level.

[0091] The decomposition result storage unit 44 stores the contribution degree for each combination of two variables included in the plurality of variables for each of one or more pieces of performance data calculated by the contribution decomposition unit 42. In this embodiment, the decomposition result storage unit 44 stores a contribution tensor C.

[0092] The output unit 46 outputs, for example, by displaying on a display device, the contribution degree for each pair of two variables included in the plurality of variables for each of one or more pieces of performance data stored in the decomposition result storage unit 44. In this embodiment, the output unit 46 outputs the value of at least one element included in the contribution tensor C stored in the decomposition result storage unit 44, or a value obtained by converting the value of at least one element included in the contribution tensor C stored in the decomposition result storage unit 44.

[0093] The output unit 46 may accept a selection of at least one piece of performance data of interest indicating a sample of interest from among the one or more pieces of performance data. The output unit 46 may also accept a selection of at least one source variable of interest to be selected as a source variable from among the plurality of variables. The output unit 46 may also accept a selection of at least one target variable of interest to be selected as a target variable from among the plurality of variables. The output unit 46 may accept the selection of the at least one piece of performance data of interest, the at least one source variable of interest, and the at least one target variable of interest in response to a user operation, or may accept the selection based on preset information.

[0094] Then, the output unit 46 outputs, as cause information, contribution degrees corresponding to all or some combinations of the selected at least one piece of note performance data, the selected at least one source variable of interest, and the selected target variable of interest. For example, the output unit 46 causes a display device or the like to display cause information in the form of a table or graph showing contribution degrees corresponding to all or some combinations of the selected at least one piece of note performance data, the selected source variable of interest, and the selected target variable of interest. Specific examples of the information displayed by the output unit 46 will be described in detail below.

[0095] FIG. 9 is a diagram showing the results of a comparison between the analysis device 30 according to this embodiment and a conventional regression model.

[0096] When analyzing changes in a specific variable using a conventional regression model, the specific variable is expressed by an equation that adds a term obtained by multiplying the actual value of another variable by a coefficient and a constant. For example, when analyzing changes in X3 of three variables, X1, X2, and X3, using a conventional regression model, X3 is expressed by equation (15).

number

[0097] However, when analyzing changes in X3 represented by such a regression model using conventional multiple regression techniques, β1, which is the regression coefficient of X1, a variable that indirectly influences X3, will be close to zero. For this reason, when analyzing changes in a specific variable using a conventional regression model, it is not possible to consider the influence of variables that indirectly influence the specific variable.

[0098] In contrast, when analyzing changes in specific variables using the analysis device 30 according to this embodiment, the specific variables are expressed by an equation that adds up terms obtained by multiplying the external noise estimates assigned to each variable by coefficients. For example, when analyzing changes in X3 of three variables X1, X2, and X3 using the analysis device 30, X3 is expressed by equation (16).

number

[0099] In this way, when analyzing a change in X3 using the analysis device 30, in addition to the contribution of X2, which is a variable that directly affects X3 (β2E2 in equation (16)), the contribution of X1, which is a variable that indirectly affects X3 (β1E1 in equation (16)), can be taken into consideration. For example, when a large external noise is applied to X1, and the effect of the external noise applied to X1 is transmitted to X3 via X2, the analysis device 30 can identify that the root cause of the change in X3 is the external noise applied to X1.

[0100] As described above, the analysis device 30 according to this embodiment can identify variables that directly and indirectly affect the fluctuations in the performance data of any of a plurality of variables.

[0101] Non-Patent Document 10 describes a technique for identifying the root cause when a target variable becomes an abnormal value. The technique in Non-Patent Document 10 calculates an anomaly score, such as a z-score, for the target variable and decomposes the anomaly score into exogenous noise for each variable using a Shapley Value. Since the technique in Non-Patent Document 10 decomposes the anomaly score, it is necessary to calculate a Shapley Value, which requires a large amount of calculation. Furthermore, since the technique in Non-Patent Document 10 converts the target variable into an anomaly score, the results vary greatly depending on which anomaly score is used. Furthermore, the technique in Non-Patent Document 10 makes it difficult to interpret the factor decomposition results.

[0102] In contrast, the analysis device 30 according to this embodiment converts target variables into anomaly scores, eliminating the need to select the type of anomaly score and enabling stable results to be obtained. Furthermore, the analysis device 30 according to this embodiment calculates contributions using linear algebraic operations, reducing the amount of calculation required, and calculating contributions for each sample and each variable makes it easy to interpret the results of factor analysis.

[0103] Furthermore, Non-Patent Documents 11 and 12 describe techniques for identifying the root cause when the target variable is 0 or 1. The techniques in Non-Patent Documents 11 and 12 assume that the target variable is a terminal node, i.e., a node with no child variables, and construct a logistic regression model using other variables as the target variable. The techniques in Non-Patent Documents 11 and 12 then use a Shapley Value to decompose the log odds of the target variable into exogenous noise. The techniques described in Non-Patent Documents 11 and 12 do not calculate the Shapley Value itself, but use the regression coefficients of the logistic regression, but the target variable must be a terminal node and a variable that takes a value of 0 or 1.

[0104] In contrast, the analysis device 30 according to the present embodiment can analyze continuous variables, and the variables to be analyzed do not need to be terminal nodes, making it possible to analyze a variety of variables. Furthermore, the analysis device 30 according to the present embodiment does not need to combine another model such as logistic regression in addition to the structural causal model, making it possible to perform analysis easily.

[0105] 10 is a diagram showing an example of the configuration of the output unit 46. For example, the output unit 46 includes a selection receiving unit 52 and a display control unit .

[0106] The selection receiving unit 52 receives a selection of at least one target data item from the one or more pieces of performance data. The selection receiving unit 52 also receives a selection of at least one target variable item from the plurality of variables. The selection receiving unit 52 also receives a selection of at least one target variable item from the plurality of variables.

[0107] The selection receiving unit 52 may receive a selection of at least one piece of performance data of interest, at least one source variable of interest, and at least one target variable of interest from the user, or may receive, for example, a preset value as the selected value.

[0108] The display control unit 54 reads out the contribution degree from the decomposition result memory unit 44 for each combination of all of the source variables and target variables of interest included in the at least one selected source variable of interest and the at least one selected target variable of interest, for each of the at least one selected performance data of interest.

[0109] Then, the display control unit 54 outputs the read contribution degrees as cause information. For example, the display control unit 54 causes the display device to display cause information in which the contribution degrees for each combination of all the source variables of interest and the target variables of interest are represented in the form of a table or graph.

[0110] FIG. 11 is a diagram illustrating a first example of cause information.

[0111] For example, the display control unit 54 may output a first example of cause information as shown in Fig. 11, which includes a table in which one of the target variable and the target variable of interest is in the rows and the other is in the columns. In this case, the display control unit 54 includes the numerical value of the corresponding contribution degree in the cell of the table.

[0112] For example, when the selection of the performance data of i=1 is accepted as the performance data of interest, the selection of the variables of j=1 to 5 is accepted as the source variables of interest, and the selection of the variables of k=1 to 5 is accepted as the target variables of interest, the display control unit 54 selects C included in the contribution tensor C. 1jk The display control unit 54 reads out the values ​​of the elements of (j=1 to 5, k=1 to 5) from the decomposition result storage unit 44. Then, the display control unit 54 reads out the values ​​of the elements of C included in the contribution tensor C. 1jk A table containing the values ​​of elements (j=1 to 5, k=1 to 5) in the cells, with j components as rows and k components as columns, as shown in FIG. 11, is output as cause information.

[0113] Such a display control unit 54 can show the user how, in the performance data of interest where i=1, the extrinsic noise (E1 to E5) given to the source variables of interest where j=1 to 5 affects the target variables of interest (X1 to X5) where k=1 to 5. This allows the user to easily confirm which extrinsic noise of the source variables of interest where j=1 to 5 is causing a change in the value of each of the target variables of interest where k=1 to 5.

[0114] FIG. 12 is a diagram illustrating a second example of the cause information.

[0115] For example, the display control unit 54 may output a second example of cause information as shown in Fig. 12, which includes a table in which one of the focused performance data and the focused source variable is in the rows and the other is in the columns. In this case, the display control unit 54 includes the numerical value of the corresponding contribution degree in the cell of the table.

[0116] For example, when the selection of performance data of i=1 to 10 is accepted as the performance data of interest, the selection of variables of j=1 to 5 is accepted as the source variables of interest, and the selection of variable of k=5 is accepted as the target variable of interest, the display control unit 54 selects C included in the contribution tensor C. ij5 The display control unit 54 reads out the values ​​of the elements of (i=1 to 10, j=1 to 5) from the decomposition result storage unit 44. Then, the display control unit 54 reads out the values ​​of the elements of C included in the contribution tensor C. ij5 A table containing the values ​​of elements (i=1 to 10, j=1 to 5) in the cells, with i components as rows and j components as columns, as shown in FIG. 12, is output as cause information.

[0117] Such a display control unit 54 can show the user how the influence of extrinsic noise (E1 to E5) given to each of the source variables of interest (j = 1 to 5) on the target variable of interest (X5) of k = 5 changes in the performance data of interest of i = 1 to 10. This allows the user to easily confirm, for each of the performance data of interest of i = 1 to 10, which extrinsic noise of the source variables of interest (j = 1 to 5) is affecting the change in the value of the target variable of interest (k = 5).

[0118] FIG. 13 is a diagram illustrating a third example of cause information.

[0119] For example, the display control unit 54 may output a third example of cause information as shown in Fig. 13, which includes a table in which one of the focused performance data and the focused target variable is in the rows and the other is in the columns. In this case, the display control unit 54 includes the corresponding contribution degree in the cell of the table.

[0120] For example, when the selection of performance data of i=1 to 10 is accepted as the performance data of interest, the selection of the variable of j=1 is accepted as the source variable of interest, and the selection of the variables of k=1 to 5 is accepted as the target variable of interest, the display control unit 54 selects C included in the contribution tensor C. i1k (i=1 to 10, k=1 to 5) element values ​​are read from the decomposition result storage unit 44. Then, the display control unit 54 reads out the C i1kA table containing the element values ​​(i=1 to 10, k=1 to 5) in the cells, with i components as rows and k components as columns, as shown in FIG. 13, is output as cause information.

[0121] Such a display control unit 54 can show the user how the influence of the extrinsic noise (E1) given to the source variable of interest (X1) of j=1 on each of the target variables of interest (X1 to X5) of k=1 to 5 changes in the performance data of interest of i=1 to 10. This allows the user to easily confirm which of the target variables of interest (k=1 to 5) is being influenced by the extrinsic noise of the source variable of interest (j=1) for each of the performance data of interest of i=1 to 10.

[0122] FIG. 14 is a diagram illustrating a fourth example of the cause information.

[0123] For example, the display control unit 54 may output a fourth example of cause information as shown in FIG. 14, which includes a graph corresponding to at least one source variable of interest drawn on a planar region. In this case, the planar region represents an index identifying the order of each of the multiple performance data of interest on the horizontal axis, which is an example of the first axis, and represents the degree of contribution on the vertical axis, which is an example of the second axis. In this case, the graph corresponding to each of the at least one source variable of interest represents the degree of contribution that external noise given to the source variable of interest corresponding to the index makes to the target variable of interest. For example, the display control unit 54 outputs a graph in which the line type or the shape of the plotted mark is changed for each source variable of interest.

[0124] Such a display control unit 54 can visually show the user, for example, how the influence of external noise given to each of at least one source variable of interest on a target variable of interest changes in accordance with changes in the index of the performance data of interest.

[0125] Furthermore, the vertical axis, which is the second axis of the planar region in the cause information of the fourth example shown in FIG. 14 , may further represent the performance value of the target variable of interest. The display control unit 54 may then output, to the planar region, cause information that further includes a graph representing the performance value of the target variable of interest for each of the selected multiple pieces of performance data of interest. The sum of the contributions of multiple source variables is the performance value of the target variable of interest. Therefore, by outputting cause information that further includes a graph representing the performance value of such a target variable of interest, the display control unit 54 can visually show the user the proportion by which the contribution of each external noise to at least one source variable of interest is resolved and how the proportion of the contribution is changing.

[0126] FIG. 15 is a diagram illustrating a fifth example of the cause information.

[0127] For example, the display control unit 54 may output a fifth example of cause information as shown in FIG. 15, which includes a pie chart or a bar graph showing the proportion of the contribution of each of a plurality of target source variables to the target performance data.

[0128] In this case, the display control unit 54 calculates the ratio of the contribution of extrinsic noise of each of the selected source variables of interest to the target variable of interest for the performance data of interest. For example, the display control unit 54 calculates, as the ratio of the contribution of extrinsic noise of each of the selected source variables of interest, the ratio of the magnitude of the contribution of extrinsic noise of the corresponding source variable of interest to the total value of the magnitude of the contribution of extrinsic noise of all of the source variables of interest.

[0129] Alternatively, the display control unit 54 calculates, as the proportion of the contribution of the exogenous noise of each of the selected source variables of interest, the proportion of the statistical value of the contribution of the exogenous noise of the corresponding source variable of interest to the target variable of interest to the total value of the statistical values ​​of the contribution of the exogenous noise of each of the source variables of interest to the target variable of interest. The statistical value of the contribution is, for example, the variance of the contribution, the standard deviation of the contribution, the sum of the squares of the contribution, or the square root of the sum of the squares of the contribution.

[0130] The display control unit 54 then generates a pie chart for the performance data of interest showing the proportion of the contribution of each of the selected source variables of interest to the target variable of interest. Alternatively, the display control unit 54 may generate a bar graph showing the proportion of the contribution of exogenous noise of each of the selected source variables of interest to the target variable of interest, drawn in a plane area where the horizontal axis represents information identifying each of the source variables of interest and the vertical axis represents the proportion.

[0131] Such a display control unit 54 can visually show the user the ratio of the degree of contribution of the external noise of each of the plurality of source variables of interest to the target variable of interest for the performance data of interest.

[0132] FIG. 16 is a diagram illustrating a sixth example of the cause information.

[0133] For example, the display control unit 54 may output a sixth example of cause information as shown in FIG. 16, which includes bar graphs corresponding to multiple source variables of interest drawn on a planar region. In this case, the planar region has a horizontal axis, which is an example of a first axis, representing information identifying external noise of the multiple source variables of interest, and a vertical axis, which is an example of a second axis, representing the degree of contribution. In this case, the bar graphs corresponding to the multiple source variables of interest represent the magnitude of the degree of contribution. The display control unit 54 may rearrange the positions of the multiple source variables of interest on the horizontal axis according to the degree of contribution or the absolute value of the degree of contribution.

[0134] Such a display control unit 54 can visually show the user the degree of contribution of exogenous noise of each of a plurality of source variables of interest to a target variable of interest for the performance data of interest.

[0135] FIG. 17 is a diagram showing a seventh example of cause information.

[0136] For example, the display control unit 54 may output a seventh example of cause information as shown in FIG. 17, which includes a waterfall graph corresponding to each of a plurality of source variables of interest drawn on a planar region. In this case, the planar region has a horizontal axis, which is an example of a first axis, representing information identifying external noise of the plurality of source variables of interest and information identifying a total value. The planar region has a vertical axis, which is an example of a second axis, representing the magnitude of the contribution.

[0137] In this case, the waterfall graph corresponding to each of the plurality of source variables of interest represents the absolute value of the degree of contribution. Note that the position of the waterfall graph corresponding to each of the plurality of source variables of interest in the vertical axis direction in the planar region is set so as to represent the increase or decrease from the adjacent variable. Note that the display control unit 54 may rearrange the positions of the plurality of source variables of interest and the total value on the horizontal axis according to the magnitude or absolute value of the degree of contribution.

[0138] The total value of the waterfall graph represents the performance value of the target variable of interest. Therefore, the display control unit 54 can visually show the user how the contribution of each of the multiple source variables of interest to the target variable of interest is decomposed for the performance data of interest.

[0139] FIG. 18 is a diagram illustrating an eighth example of the cause information.

[0140] For example, the display control unit 54 may output an eighth example of cause information as shown in FIG. 18, which includes a graph corresponding to each of at least one target variable of interest drawn on a planar region. In this case, the planar region represents an index identifying the order of each of the multiple performance data of interest on the horizontal axis, which is an example of the first axis, and represents the magnitude of the contribution on the vertical axis, which is an example of the second axis. In this case, the graph corresponding to each of at least one target variable of interest represents the contribution of exogenous noise given to the source variable of interest corresponding to the index to the target variable of interest. For example, the display control unit 54 outputs a graph in which the line type or the shape of the plotted mark is changed for each target variable of interest.

[0141] Such a display control unit 54 can visually show the user, for example, how the influence of external noise given to each of at least one target variable of interest on a source variable of interest changes in accordance with changes in the index of the performance data of interest.

[0142] FIG. 19 is a diagram illustrating a ninth example of cause information.

[0143] 19, which includes a table with one of the source variable of interest and the target variable of interest as rows and the other as columns. In this case, the cells in the table include a numerical value representing the degree of contribution that external noise given to the corresponding source variable of interest makes to the corresponding target variable of interest, or a value obtained by converting a numerical value representing the degree of contribution that external noise given to the corresponding source variable of interest makes to the corresponding target variable of interest.

[0144] Furthermore, in addition to or instead of a numerical value, each cell in the table includes a background image that is highlighted according to the magnitude of the contribution of the corresponding source variable of interest to the corresponding target variable of interest. That is, the display control unit 54 displays the table as a heat map by changing the degree of highlighting of the cell according to the magnitude of the contribution within the cell. For example, the display control unit 54 changes the color or density of the cell according to the magnitude of the contribution.

[0145] This allows the display control unit 54 to visually show the user which combination of the noted source variable and the noted target variable has a large degree of contribution for the noted performance data.

[0146] FIG. 20 is a diagram showing a first display example of a causal graph and cause information.

[0147] The display control unit 54 may cause the display device to display image information representing a causal graph such as that shown in Fig. 20. In this case, the selection receiving unit 52 selects, in response to a user operation such as a click, a variable corresponding to the operated node among multiple nodes included in the causal graph as an attention source variable or an attention target variable. Then, the display control unit 54 causes the display device to display the cause information of the first to ninth examples based on the selected attention source variable or attention target variable.

[0148] The display control unit 54 may highlight and display the node selected by the user in the causal graph. For example, the display control unit 54 may display the node selected by the user with a thicker line than the other nodes or in a different color from the other nodes. The display control unit 54 may also display the node selected as the source variable of interest and the node selected as the target variable of interest in a distinguished manner.

[0149] The display control unit 54 can visually display the cause information together with the causal graph to the user, thereby enabling the user to efficiently analyze the cause or the analysis results of the cause.

[0150] FIG. 21 is a diagram showing a second display example of the causal graph and the cause information.

[0151] When image information representing a causal graph is displayed on a display device, the display control unit 54 selects, in response to a user operation such as a click, a variable corresponding to an operated node among multiple nodes included in the causal graph as a target variable of interest. In this case, the display control unit 54 selects, as source variables of interest, each of multiple variables corresponding to multiple nodes included in the causal graph.

[0152] The display control unit 54 may then highlight each of the multiple nodes according to the contribution rate when the corresponding variable is selected as the source variable of interest. For example, the display control unit 54 changes the size, color, thickness, etc. of the node according to the contribution rate. The display control unit 54 may also change the size, color, thickness, etc. of the node according to the proportion of the contribution rate. For example, the display control unit 54 may change the size, color, thickness, etc. of the node according to the proportion of the magnitude of the contribution rate, the proportion of the variance of the contribution rate, the proportion of the standard deviation, the proportion of the sum of squares of the contribution rate, or the proportion of the square root of the sum of squares. For example, the display control unit 54 may also annotate each of the multiple nodes with the contribution rate when the corresponding variable is selected as the source variable of interest.

[0153] The display control unit 54 can visually show the user the contribution of each of a plurality of source variables of interest to the target variable of interest in the exogenous noise, along with a causal graph, thereby enabling the user to efficiently perform analysis of the cause or the results of the cause analysis.

[0154] FIG. 22 is a diagram showing a third display example of the causal graph and the cause information.

[0155] When image information representing a causal graph is displayed on a display device, the display control unit 54 may display a pie chart corresponding to each of multiple nodes included in the causal graph. When the corresponding variable is set as a target variable of interest, the pie chart represents the proportion of the contribution of external noise to the source variable of interest when the corresponding variable itself and variables upstream of the corresponding variable are set as source variables of interest. In this case, the display control unit 54 displays a pie chart with the contribution of its own external noise set to 100% for the most upstream node, i.e., for a node with no parent node. The proportion of contribution is the same as that of the cause information in the fifth example described with reference to FIG. 15.

[0156] This allows the display control unit 54 to visually show the user how the contribution ratio when each of multiple variables is set as a target variable of interest changes according to the structure of the structural causal model.

[0157] 23 is a diagram showing a modified example of the information processing system 10 according to the embodiment. The information processing system 10 according to the modified example further includes an abnormal change detection device 80.

[0158] The analysis device 30 periodically acquires one or more pieces of performance data including performance values ​​of multiple variables, for example, as a routine operation, and outputs cause information. Instead of or in addition to such processing, the analysis device 30 according to the modified example acquires one or more pieces of performance data including performance values ​​of multiple variables, as a non-routine operation, in response to an instruction from the abnormal change detection device 80, and outputs cause information.

[0159] The abnormal change detection device 80 acquires one or more pieces of performance data, etc., and detects that an abnormality or a possible abnormality has occurred in the target system 20, or that the state of the target system 20 has changed. When the abnormal change detection device 80 detects a change in the state of the target system 20, it issues an operational instruction to the analysis device 30. For example, the abnormal change detection device 80 may determine the state of the target system 20 by comparing the quantile and standard deviation of the actual values ​​of one of the multiple variables to a preset threshold. Furthermore, for example, the abnormal change detection device 80 may determine the state of the target system 20 by testing the mean and standard deviation of the actual values ​​of the monitored variables. Furthermore, for example, the abnormal change detection device 80 may make an individual determination for each of two or more monitored variables among the multiple variables, and determine that the state of the target system 20 has become abnormal, changed, or both abnormal and changed when an abnormal change is detected in one or a predetermined number of the two or more monitored variables.

[0160] Furthermore, the abnormal change detection device 80 may perform multivariate analysis on two or more monitored variables among the plurality of variables to determine the state of the target system 20. For example, the abnormal change detection device 80 may determine the state of the target system 20 using a method based on Hotelling, k-NN, SVM, CAE, a graphical model, a density ratio, or the like for two or more monitored variables.

[0161] By being equipped with such an abnormal change detection device 80, the information processing system 10 can assist in the task of identifying the abnormalities, changes, or factors that caused the abnormalities and changes detected by the abnormal change detection device 80.

[0162] Furthermore, when the target system 20 is a system for manufacturing a product, the target system 20 uses a control chart to monitor and control the quality characteristics of the product. The control chart includes information that defines the control range or specification range of values ​​that represent the quality characteristics of the product.

[0163] In this case, the multiple variables include the quality characteristics of the products defined in the control chart as variables, and the structural causal model set in the analysis device 30 includes the quality characteristics of the products defined in the control chart as variables.

[0164] The abnormal change detection device 80 causes the analysis device 30 to execute processing when the actual value of a variable representing a quality characteristic of a product falls outside the control range or specification range. When the quality characteristic of a product falls outside the control range or specification range shown on the control chart, the analysis device 30 performs root cause analysis using the variable corresponding to the quality characteristic that falls outside the control range or specification range as the target variable. This allows the analysis device 30 to identify the underlying cause of the quality characteristic falling outside the control range or specification range, such as raw materials, processes, and quality characteristics. This allows the user to combine the cause information output from the analysis device 30 with the control chart to investigate the cause and consider countermeasures.

[0165] (Hardware configuration of the analysis device 30, etc.) Fig. 24 is a diagram showing an example of the hardware configuration of the analysis device 30. The analysis device 30 is realized by, for example, an information processing device having the hardware configuration shown in Fig. 24. The analysis device 30 includes a CPU (Central Processing Unit) 901, a RAM (Random Access Memory) 902, a ROM (Read Only Memory) 903, a storage device 904, and a communication interface device 905. These components are connected via a bus.

[0166] The CPU 901 is one or more processors that execute arithmetic processing, control processing, etc. according to a program. The CPU 901 uses a predetermined area of ​​the RAM 902 as a working area and executes various processes in cooperation with programs stored in the ROM 903, the storage device 904, etc.

[0167] The RAM 902 is a memory such as an SDRAM (Synchronous Dynamic Random Access Memory), and functions as a work area for the CPU 901. The ROM 903 is a memory that stores programs and various information in a non-rewritable manner.

[0168] The storage device 904 is a device that writes and reads data to a semiconductor storage medium such as a flash memory, or a magnetically or optically recordable storage medium, etc. The storage device 904 writes and reads data to the storage medium in response to control from the CPU 901. The communication interface device 905 communicates with external devices via a network in response to control from the CPU 901.

[0169] A program executed by the information processing device causes the information processing device to function as the analysis device 30. This program is loaded onto the RAM 902 by the CPU 901 (processor) and executed.

[0170] In addition, the program executed by the information processing device is provided as a file in a format that can be installed on the information processing device or in an executable format, recorded on a recording medium that can be read by the information processing device, such as a CD-ROM, flexible disk, CD-R, or DVD (Digital Versatile Disk).

[0171] This program may also be configured to be stored on a computer connected to a network such as the Internet and provided by being downloaded via the network. This program may also be configured to be provided or distributed via a network such as the Internet. The program executed by the analysis device 30 may also be configured to be provided by being pre-installed in the ROM 903 or the like.

[0172] The program for causing the information processing device to function as the analysis device 30 includes, for example, an actual data acquisition module, a model acquisition module, an external noise estimation module, a contribution decomposition module, and an output module. When executed by the CPU 901, the respective modules are loaded into the RAM 902, causing the CPU 901 to function as the actual data acquisition unit 32, the model acquisition unit 36, the external noise estimation unit 40, the contribution decomposition unit 42, and the output unit 46. If the CPU 901 has multiple processors, these units may be divided among the multiple processors. Note that these components may be partially or entirely implemented by hardware. Furthermore, the program causes the RAM 902 and the storage device 904 to function as the actual data storage unit 34, the model storage unit 38, and the decomposition result storage unit 44.

[0173] Although several embodiments of the present invention have been described, these embodiments are presented as examples and are not intended to limit the scope of the invention. These novel embodiments can be embodied in various other forms, and various omissions, substitutions, and modifications can be made without departing from the spirit of the invention. These embodiments and their modifications are included within the scope and spirit of the invention, and are also included in the scope of the invention and its equivalents as defined in the claims. [Explanation of symbols]

[0174] 10 Information Processing Systems 20 Target System 30 Analyzer 32 Performance data acquisition section 34 Performance data storage unit 36 Model Acquisition Department 38 Model Memory Unit 40 External noise estimation unit 42 Contribution decomposition part 44 Decomposition result storage section 46 Output section 80 Abnormal Change Detection Device

Claims

1. calculating, for each of the one or more pieces of performance data, a plurality of external noise estimated values ​​corresponding to the plurality of variables based on one or more pieces of performance data each including a plurality of performance values ​​corresponding to a plurality of variables and a structural causal model representing a causal relationship between the plurality of variables, each of the plurality of external noise estimated values ​​representing an estimated value of an influence of external noise different from an influence from the plurality of variables on a corresponding variable among the plurality of variables; Based on the structural causal model and the plurality of exogenous noise estimates for each of the one or more pieces of actual data, a contribution is generated for each combination of two variables included in the plurality of variables, the contribution representing the magnitude of the influence that the exogenous noise imparted to a source variable, which is one of the two variables, has on a target variable, which is the other of the two variables. Processing section An information processing device comprising:

2. The structural causal model is expressed using an adjacency matrix that indicates the magnitude of the influence from one variable to the other variable for each combination of two variables included in the plurality of variables. The information processing device according to claim 1 .

3. the processing unit calculates an extrinsic noise matrix by multiplying an achievement data matrix including the one or more achievement data by a matrix obtained by subtracting the adjacent matrix from a unit matrix; The external noise matrix includes the plurality of external noise estimation values ​​for each of the one or more pieces of actual data. The information processing device according to claim 2 .

4. When a first variable of the plurality of variables in first performance data of the one or more performance data is set as the target variable and a second variable of the plurality of variables is set as the source variable, the contribution to a combination of the two variables is expressed as: is a value of a term including an external noise estimate value representing the external noise assigned to the second variable, among a plurality of terms included in the linear sum equation; the linear sum formula is a formula for performing a product-sum operation on a row or a column in the external noise matrix corresponding to the first performance data and a row or a column in a coefficient matrix corresponding to the first variable, The coefficient matrix is ​​the inverse matrix or generalized inverse matrix of the matrix obtained by subtracting the adjacent matrix from the identity matrix. The information processing device according to claim 3 .

5. The processing unit generates a third-order tensor including an i-component, a j-component, and a k-component; The i component corresponds to each of the one or more pieces of performance data, The j component corresponds to each of the plurality of variables, the k components correspond to each of the plurality of variables; The elements included in the third-order tensor represent the degree of contribution of the one or more pieces of performance data identified by the i component to a variable identified by the j component among the plurality of variables as the source variable and a variable identified by the k component among the plurality of variables as the target variable. The information processing device according to claim 1 .

6. The adjacency matrix is ​​estimated using a causal search algorithm based on the one or more pieces of performance data. The information processing device according to claim 2 .

7. The adjacency matrix is ​​estimated based on the one or more historical data using known causal structure and covariance structure analysis. The information processing device according to claim 2 .

8. The adjacency matrix is ​​estimated based on the one or more pieces of performance data, using information on the presence or absence of causality between variables and a causality search algorithm. The information processing device according to claim 2 .

9. The processing unit Accepting selection of a target variable of interest from among the one or more pieces of performance data, a source variable of interest from among the plurality of variables, and a target variable of interest from among the plurality of variables. The information processing device according to claim 1 .

10. The processing unit For a target performance data item of the one or more performance data items, output cause information representing the degree of contribution, with the target variable of the target variable being the target source variable of the plurality of variables and the target variable being the target variable. The information processing device according to claim 1 .

11. The cause information includes a table or graph showing the contributions corresponding to all combinations of the source variable of interest and the target variable of interest included in at least one source variable of interest and at least one target variable of interest for at least one performance data of interest. The information processing device according to claim 10.

12. the cause information includes a graph corresponding to each of the at least one source variable of interest in a planar region whose position is specified by a first axis representing an index that identifies each of the one or more performance data and a second axis representing the degree of contribution; The graph corresponding to each of the at least one source variable of interest represents the degree of contribution of the external noise given to the source variable of interest corresponding to the index to the target variable of interest. The information processing device according to claim 10.

13. the second axis further represents the actual value of the target variable of interest; The cause information further includes a graph plotted on the planar region, the graph representing the actual value of the target variable of interest corresponding to the index. The information processing device according to claim 12.

14. the cause information includes a pie chart or a bar graph representing the proportion of the contribution of each of the plurality of source variables of interest for the performance data of interest; The contribution ratio represents the ratio of the magnitude of the contribution of the external noise of the corresponding source variable of interest to the target variable of interest to the total value of the magnitude of the contribution of the external noise of each of the plurality of source variables of interest to the target variable of interest, or the ratio of the statistical value of the contribution of the external noise of the corresponding source variable of interest to the target variable of interest to the total value of the statistical value of the contribution of the external noise of each of the plurality of source variables of interest to the target variable of interest. The information processing device according to claim 10.

15. The cause information includes a bar graph or a waterfall graph showing the contribution of each of the plurality of source variables of interest to the performance data of interest. The information processing device according to claim 10.

16. the cause information includes a graph corresponding to each of the at least one target variable of interest in a planar region whose position is specified by a first axis representing an index that identifies each of the one or more performance data and a second axis representing the degree of contribution; The graph corresponding to each of the at least one target variable of interest represents the degree of contribution of the exogenous noise given to the source variable of interest corresponding to the index to the corresponding target variable of interest. The information processing device according to claim 10.

17. the cause information includes a table in which one of the columns or rows indicates the source variables of interest and the other indicates the target variables of interest for the performance data of interest; The cells in the table include a numerical value representing the degree of contribution of the external noise given to the corresponding source variable of interest to the corresponding target variable of interest, and / or a background image highlighted according to the degree of contribution of the corresponding source variable of interest to the corresponding target variable of interest. The information processing device according to claim 10.

18. The processing unit displaying a causal graph representing the structural causal model; each of a plurality of nodes included in the causal graph represents one of the plurality of variables; A variable corresponding to a node operated by a user is selected as the source variable of interest or the target variable of interest from among the plurality of nodes included in the causal graph. The information processing device according to claim 10.

19. The processing unit displaying a causal graph representing the structural causal model; each of a plurality of nodes included in the causal graph represents one of the plurality of variables; In response to a user selecting any one of the plurality of nodes included in the causal graph as the target variable of interest, each of the plurality of nodes is highlighted according to the contribution when the corresponding variable is selected as the source variable of interest. The information processing device according to claim 10.

20. The processing unit displaying a causal graph representing the structural causal model; each of a plurality of nodes included in the causal graph represents one of the plurality of variables; displaying an image including a pie chart or a bar graph in correspondence with each of the plurality of nodes; the pie chart or the bar graph represents the proportion of the contribution of each of a plurality of source variables of interest when the corresponding variable or a variable upstream of the corresponding variable in the causal graph is defined as the plurality of source variables of interest; The contribution ratio represents the ratio of the magnitude of the contribution of the external noise of the corresponding source variable of interest to the target variable of interest to the total value of the magnitude of the contribution of the external noise of each of the plurality of source variables of interest to the target variable of interest, or the ratio of the statistical value of the contribution of the external noise of the corresponding source variable of interest to the target variable of interest to the total value of the statistical value of the contribution of the external noise of each of the plurality of source variables of interest to the target variable of interest. The information processing device according to claim 10.

21. An information processing device according to any one of claims 1 to 20; an abnormal change detection device that detects an abnormality or change in the state of the target system that outputs the one or more pieces of performance data; Equipped with When the abnormal change detection device detects an abnormality or a change in the state of the target system, the abnormal change detection device causes the information processing device to generate the contribution degree. Information processing system.

22. the target system is a system for manufacturing a product, the plurality of variables includes a quality characteristic of the product as a variable; When an actual value of a variable representing a quality characteristic of the product falls outside a control range or a specification range defined by a control chart generated in advance, the abnormal change detection device causes the information processing device to generate the degree of contribution.

22. The information processing system according to claim 21.

23. An information processing method executed by an information processing device, the information processing device calculates, for each of the one or more pieces of performance data, a plurality of external noise estimated values ​​corresponding to the plurality of variables, based on one or more pieces of performance data including a plurality of performance values ​​each corresponding to a plurality of variables and a structural causal model representing a causal relationship between the plurality of variables, and each of the plurality of external noise estimated values ​​represents an estimated value of an influence of external noise different from an influence from the plurality of variables on a corresponding variable among the plurality of variables; The information processing device generates, for each combination of two variables included in the plurality of variables, a contribution representing the magnitude of the influence that the extrinsic noise given to a source variable, which is one of the two variables, has on a target variable, which is the other of the two variables, for each of the one or more performance data, based on the structural causal model and the plurality of extrinsic noise estimated values ​​for each of the one or more performance data. Information processing methods.

24. Computer, calculating, for each of the one or more pieces of performance data, a plurality of external noise estimated values ​​corresponding to the plurality of variables based on one or more pieces of performance data each including a plurality of performance values ​​corresponding to a plurality of variables and a structural causal model representing a causal relationship between the plurality of variables, each of the plurality of external noise estimated values ​​representing an estimated value of an influence of external noise different from an influence from the plurality of variables on a corresponding variable among the plurality of variables; Based on the structural causal model and the plurality of exogenous noise estimates for each of the one or more pieces of actual data, a contribution is generated for each combination of two variables included in the plurality of variables, the contribution representing the magnitude of the influence that the exogenous noise imparted to a source variable, which is one of the two variables, has on a target variable, which is the other of the two variables. Processing section A program to function as a

Images