JP7728199B2 - Analytical device, analytical method, and analytical program - Google Patents

Description

The present invention relates to an analytical device, an analytical method, and an analytical program.

A technique known as imitation learning is known in which a machine learning model learns human behavior and then uses that model to teach actions to humans or robots.

For example, in imitation learning at chemical plants, machine learning models may be used in which multiple sensor values are input as features and information about equipment operation is output.

In addition, a method has been proposed for analyzing causal relationships between features.

For example, a method has been proposed that can use numerical calculations to derive a directed acyclic graph (DAG), which is expected to represent causal relationships (see, for example, Non-Patent Document 1).

In addition, for example, a method has been proposed that combines the method described in Non-Patent Document 1 with a type of neural network, MLP (Multilayer Perceptron), to capture non-linear causal relationships using DAG (see, for example, Non-Patent Document 2).

Zheng, Xun, et al. "DAGs with NO TEARS: Continuous Optimization for Structure Learning." Advances in Neural Information ProcessingSystems 31, 2018. Zheng, Xun, et al. "Learning sparse nonparametric dags." International Conference on Artificial Intelligence and Statistics. PMLR, 2020.

However, conventional methods for analyzing causal relationships have the problem that it can be difficult to properly reflect known causal relationships in the results of analyzing causal relationships between data.

Here, the presence or absence of causality between some data may be known to the user as prior knowledge. Note that the data may be, for example, features input into a machine learning model.

On the other hand, with conventional analysis methods, causal relationships between data are uniquely inferred using data and algorithms, making it difficult to reflect prior knowledge in the analysis results.

To solve the above-mentioned problems and achieve the objectives, the analysis device is characterized by having: a setting unit that sets fixed values corresponding to the causal relationship to the weights of nodes in the input layer of a neural network, among multiple nodes to which multiple feature quantities are input, where the feature quantities have a pre-specified causal relationship with the result data; an update unit that updates the weights of nodes included in the neural network that have not been set to fixed values by the setting unit so that the output of the neural network approaches the result data; and a generation unit that generates information indicating the causal relationship between the multiple feature quantities and the result data based on the weights of the nodes in the input layer.

According to the present invention, known causal relationships can be appropriately reflected in the analysis results of causal relationships between feature quantities.

FIG. 1 is a diagram illustrating a plant operation system. FIG. 2 is a diagram illustrating an example of the configuration of a server according to the first embodiment. FIG. 3 is a diagram illustrating an example of feature amount information. FIG. 4 is a diagram illustrating an example of the configuration of a model. FIG. 5 is a diagram for explaining a setting method when there is causality between data. FIG. 6 is a diagram for explaining a setting method when there is no causality between data. FIG. 7 is a diagram for explaining a setting method when the causality between data is unknown. FIG. 8 is a diagram illustrating an example of a causal graph. FIG. 9 is a flowchart showing the flow of the analysis process according to the first embodiment. FIG. 10 is a flowchart showing the flow of the setting process according to the first embodiment. FIG. 11 is a diagram illustrating an example of the configuration of a computer that executes an analysis program.

Embodiments of the analytical device, analytical method, and analytical program according to the present application are described in detail below with reference to the accompanying drawings. Note that the present invention is not limited to the embodiments described below.

[First embodiment]
[Configuration of the first embodiment]
First, a plant operation system will be described with reference to Fig. 1. The plant operation system 1 is a system for managing and controlling the production process of products in a plant. The plant includes a chemical plant for producing chemical products.

As shown in Figure 1, it has a server 10, a terminal device 20, and a plant system 30.

Furthermore, the server 10, terminal device 20, and plant system 30 are connected to each other so that they can communicate data with each other via a network N. For example, the network N is the Internet or an intranet.

Server 10 performs processing related to imitation learning. Server 10 executes learning processing of a model for imitation learning (hereinafter referred to as the imitation learning model) and inference processing using the imitation learning model.

For example, the server 10 acquires a history of sensor values and corresponding operation details at each time from the plant system 30. The sensor values and operation details correspond to the explanatory variables and objective variables of the imitation learning model, respectively.

In this case, the server 10 performs the learning process for the imitation learning model using the sensor values and operation details acquired from the plant system 30 as training data.

In addition, the server 10 inputs the sensor values acquired from the plant system 30 into the trained imitation learning model and performs inference processing to obtain the operation details.

The server 10 provides the user with the operation details obtained through the inference process. The user can operate the equipment in the plant system 30 according to the provided operation details.

For example, the user can operate the device to set the temperature inside the device, the pressure inside the device, the target production volume for the production process, the amount of raw materials to be fed into the device, etc.

The terminal device 20 is an information processing device such as a personal computer, tablet terminal, or smartphone.

The plant system 30 may include equipment used in the production process and a distributed control system (DCS). For example, the equipment may be a reactor, a cooler, a gas-liquid separator, etc.

Furthermore, the server 10 performs a causal analysis process that analyzes the causal relationships between variables in the imitation learning model. The server 10 is an example of an analysis device. The process flow for analyzing causal relationships will be explained using Figure 1.

Note that the variables of the imitation learning model include both explanatory variables and target variables. In the following explanation, the variables of the imitation learning model may be referred to as features or simply as data.

First, as shown in FIG. 1, the server 10 accumulates data acquired from the plant system 30 (step S1). For example, the server 10 acquires and accumulates sensor values such as temperature, concentration, pressure, and flow rate for each time period.

Next, the server 10 acquires prior knowledge regarding the causal relationships between the features from the user (step S2).

For example, prior knowledge may be provided to the server 10 as a list of feature pairs that the user believes to be causal and the direction of causality, or a list of feature pairs that the user believes to be non-causal.

In response to the analyst's operation, the terminal device 20 instructs the server 10 to execute the causal analysis process (step S3). The terminal device 20 also obtains the causal analysis results from the server 10.

The server 10 adds and deletes causal relationships between features based on prior knowledge, and then calculates the causal relationships according to an algorithm. For example, the server 10 outputs a DAG (Directed Acyclic Graph) generated based on the calculated causal relationships as the causal analysis result.

The server 10 will be described in detail using Figure 2. Figure 2 is a diagram showing an example of the configuration of a server according to the first embodiment.

Here, we will explain the configuration when server 10 performs causal analysis processing. We will not explain the configuration when server 10 performs learning processing and inference processing.

As shown in FIG. 2, the server 10 has a communication unit 11, a memory unit 12, and a control unit 13.

The communication unit 11 communicates data with other devices via a network. For example, the communication unit 11 is a network interface card (NIC).

The storage unit 12 is a storage device such as an HDD (Hard Disk Drive), SSD (Solid State Drive), or optical disc. The storage unit 12 may also be a data-rewritable semiconductor memory such as RAM (Random Access Memory), flash memory, or NVSRAM (Non-Volatile Static Random Access Memory).

The memory unit 12 stores the OS (Operating System) and various programs executed on the server 10. The memory unit 12 stores a history DB 121 and model information 122.

History DB 121 contains information including history provided by plant system 30. Figure 3 shows an example of the history DB.

As shown in Figure 3, history DB121 includes time, temperature, concentration, pressure, flow rate, and set value.

Temperature, concentration, pressure, and flow rate correspond to the explanatory variables of the imitation learning model. The set value corresponds to the objective variable of the imitation learning model. Note that the features that are the subject of causal analysis may include either explanatory variables or objective variables.

Furthermore, for example, the temperature, concentration, pressure, and flow rate are sensor values from sensors installed at various locations in the plant system 30. The time is a timestamp indicating the time when the temperature, concentration, pressure, and flow rate were acquired.

The setting value is a value set by operation from the terminal device 20. The setting value may be a normalized value of the actually set value. The setting value is a value used to specify the operation content of the device.

For example, Figure 3 shows that at the time "2021/11/5 13:30:01", the temperature is "40°C", the concentration is 3.5%, the pressure is "501 hPa", the flow rate is "12 m ³ /s", and the set value is "0.2".

Model information 122 is information such as parameters for constructing a model for causal analysis. For example, if the model is a neural network, model information 122 is the weights and biases of each layer.

The control unit 13 controls the entire server 10. The control unit 13 is, for example, an electronic circuit such as a CPU (Central Processing Unit), MPU (Micro Processing Unit), or GPU (Graphics Processing Unit), or an integrated circuit such as an ASIC (Application Specific Integrated Circuit) or FPGA (Field Programmable Gate Array).

The control unit 13 also has internal memory for storing programs that define various processing procedures and control data, and executes each process using the internal memory. The control unit 13 also functions as various processing units when various programs are run. For example, the control unit 13 has a setting unit 131, a calculation unit 132, an update unit 133, and a generation unit 134.

Here, we will explain an overview of the causal analysis process in this embodiment and the model used in the causal analysis process.

First, Non-Patent Document 1 describes how causal relationships between data are expressed using DAGs.

Non-Patent Document 2 also describes relaxing the problem formulation (equation (1)) used to represent causal relationships using DAGs to a problem formulation (equation (2)) using a multilayer perceptron, a type of neural network.

Here, the data to be analyzed is x _j (1, ..., d). Furthermore, X is a set of data x ₁ , ..., x _d . If x _j (1, ..., d) is a vector with n elements, X can be expressed as an n x d matrix.

Function l outputs the deviation between vectors (for example, the sum of squares in the least squares method, or the negative log-likelihood). Function L is an error function. Function G is a function that represents a DAG.

In equation (2), d multilayer perceptron MLPs are used, each corresponding to data x _j (1, . . . , d), where θ _j (1, . . . , d) are parameters of each MLP.

A _j ⁽¹⁾ is the weight of the input layer (first layer) of the multilayer perceptron corresponding to x _j . λ is a constant that is preset as a hyperparameter. That is, the second term in equation (2) represents the L1 constraint on the weight of the first layer of the MLP.

Also, h(W(θ)) = 0 is an equality constraint that induces a DAG, and means that all necessary components (e.g., diagonal components) must be degenerated to 0 so that the matrix W(θ) represents a DAG. W(θ) is an adjacency matrix that represents the weights between specified nodes.

Here, the (i, j) element of the adjacency matrix W(θ) represents the causality from data x _i to data x _j (x _i →x _j ). For example, w ₄₁ represents the causality from data x ₄ to data x ₁ .

In particular, when the (i, j) element of the adjacency matrix W(θ) is 0, it means that there is no causality from data x _i to data x _j . Conversely, when the (i, j) element of the adjacency matrix W(θ) is non-zero, it means that there is causality from data x _i to data x _j .

Constraint h(W(θ)) can be said to be a constraint that, as a result of MLP learning, the causal graph (network) corresponding to the adjacency matrix W(θ) becomes directed acyclic. Specifically, constraint h(W(θ)) is a constraint that the adjacency matrix W(θ) is a lower triangular matrix or an upper triangular matrix, with diagonal elements of 0.

In this case, due to the DAG assumption, there is no causality to each variable itself (causality from data x _i to data x _i ), so the initial values of the diagonal elements (w _ij when i = j) of the adjacency matrix W(θ) are set to 0.

Additionally, the initial values of the components other than the diagonal components (w _ij when i≠j) are set to arbitrary values, and as learning progresses, some of the components degenerate to 0 due to constraints.

In an embodiment, the server 10 reflects the causal relationships specified in advance by the user or the like as prior knowledge in the problem setting of equation (2).

The model information 122 includes information on a model for performing causal analysis processing in accordance with the problem setting of equation (2). Note that an example where d=4, that is, the data is _x1 , _x2 , _x3 , and x4 _, will be described here.

The data x ₁ , x ₂ , x ₃ , and x ₄ may be variables stored in the history DB. Furthermore, the data x ₁ , x ₂ , x ₃ , and x ₄ can be referred to as feature quantities.

Based on the model information 122, the server 10 constructs a model as shown in Figure 4. Figure 4 shows an example of the model configuration.

As shown in Figure 4, the model includes MLP51, MLP52, MLP53, and MLP54.

Each MLP corresponds to a causal analysis in which a variable in the input layer is the cause and a variable specified in the output layer is the result. MLP51, MLP52, MLP53, and MLP54 are associated with causal analyses in which data _x1 , _x2 , _x3 , and _x4 are the results, respectively.

For example, data x ₂ , x ₃ , and x ₄ are input to the MLP 51. Then, the MLP 51 is optimized so that the value output from the output layer approaches the resultant data x ₁ .

In FIG. 4, a node corresponding to data _x1 is arranged in the input layer of the MLP, but this node is for the purpose of explanation and does not actually exist or is treated as having a weight of 0.

Like MLP51, MLP52, MLP53, and MLP54 have similar configurations depending on the associated data.

For example, data x ₁ , x ₃ , and x ₄ are input to the MLP 52. Then, the MLP 52 is optimized so that the value output from the output layer approaches x ₂ .

For example, data x ₁ , x ₂ , and x ₄ are input to the MLP 53. Then, the MLP 53 is optimized so that the value output from the output layer approaches x ₃ .

For example, data _x1 , _x2 , and _x3 are input to the MLP 54. Then, the MLP 54 is optimized so that the value output from the output layer approaches _x4 .

Each part of the control unit 13 performs processing to analyze the causal relationships between the data x ₁ , x ₂ , x ₃ , and x ₄ using the model shown in FIG.

The setting unit 131 is a node in the input layer of the neural network, and among the multiple nodes to which each of the multiple feature quantities is input, sets a fixed value corresponding to the causal relationship to the weight of the node to which the feature quantity having a pre-specified causal relationship with the resulting data is input.

That is, the setting unit 131 sets fixed values for the weights from the input layer nodes of MLP51, MLP52, MLP53, and MLP54 to each node in the hidden layer based on the user's prior knowledge.

Each MLP is an example of a neural network. Data _x1 , _x2 , _x3 , and _x4 are both feature quantities and result data. For example, in MLP 52, data _x1 , _x3 , and _x4 are feature quantities, and data _x2 is result data.

5 is a diagram for explaining a setting method when there is a causal relationship between data items. Here, it is assumed that there is a causal relationship between data item _x3 and data item _x2 .

In this case, the setting unit 131 sets the weight from the node in the input layer of the MLP 52 to which the data _x3 is input to each node in the hidden layer to 1/ _√m1 , where _m1 is the number of nodes in the hidden layer connected to the input layer.

5, the number of nodes in the hidden layer connected to the input layer is 5. In this case, the setting unit 131 sets [A ₂ ⁽¹⁾ ] _.,3 = 1/√5.

Here, [A _j ⁽¹⁾ ] _k,i is the weight from the i-th node in the first layer (input layer) of the MLP associated with the causal analysis with data x _j as the result to the k-th node in the second layer (first hidden layer).

In other words, [A _j ⁽¹⁾ ] _k,i is a three-dimensional array having as its element a i _{,k,j (} ¹⁾ , which is the weight from the i-th node in the input layer to the k-th node in the first hidden layer of the MLP associated with the causal analysis with data x j as the result.

Also, "." means any value. Therefore, [A _j ⁽¹⁾ ] _{., i} is the weight from the i-th node in the first layer (input layer) of the MLP associated with the causal analysis with data x _j as the result to any node in the second layer (first hidden layer).

In this way, the setting unit 131 sets a fixed value to the weight of a node to which a feature that is specified to have a causal relationship with the result data is input, among the multiple feature amounts. 1/√m ₁ is an example of the fixed value.

6 is a diagram illustrating a setting method when there is no causal relationship between data items. Here, it is assumed that there is no causal relationship between data item _x3 and data item _x2 .

In this case, the setting unit 131 sets the weights from the nodes in the input layer of the MLP 52 to which the data _x3 is input to each node in the hidden layer to 0.

6, the setting unit 131 sets the weight as follows: [A ₂ ⁽¹⁾ ] _.,3 = 0. Note that setting the weight to 0 means deleting the connection between the nodes.

In this way, the setting unit 131 sets the weight of a node to 0 for a feature that is specified as having no causal relationship with the result data, among multiple feature values.

7 is a diagram for explaining a setting method when the causality between data is unknown. Here, it is assumed that the causal relationship from data _x3 to data _x2 is not specified.

In this case, the setting unit 131 does not set fixed values to the weights from the nodes in the input layer of the MLP 52 to which the data _x3 is input to each node in the hidden layer.

At this time, in the process by the update unit 133 described later, the weight of the node is updated within the range of [A ₂ ⁽¹⁾ ] _{., 3ε} [0, 1/√5].

The setting unit 131 also sets fixed values for MLP51, MLP53, and MLP54 using the method described in Figures 5, 6, and 7.

In other words, for each of a plurality of neural networks to which each of a plurality of feature quantities is associated as result data, the setting unit 131 sets a fixed value corresponding to the causal relationship to the weight of the node in the input layer to which each of the plurality of feature quantities is input, among a plurality of nodes to which each of the plurality of feature quantities is input, that is, a node to which a feature quantity having a pre-specified causal relationship with the associated result data is input.

The calculation unit 132 performs calculations of the MLPs. That is, the calculation unit 132 calculates MLP(X, θ ₁ ), MLP(X, θ ₂ ), MLP(X, θ ₃ ), and MLP(X, θ ₄ ).

The update unit 133 performs learning processing for each MLP. That is, the update unit 133 updates the parameters of each MLP based on the calculation results by the calculation unit 132. For example, the update unit 133 updates the parameters using the backpropagation algorithm.

The update unit 133 minimizes the error function shown in equation (2). In particular, when data _x1 , _x2 , _x3 , and _x4 are targeted, the first term of the error function in equation (2) is expressed as L' in equation (3).

The update unit 133 does not update weights for which fixed values have been set. In other words, the update unit 133 updates the weights of the nodes included in the neural network for which fixed values have not been set by the setting unit 131, so that the output of the neural network approaches the resulting data.

If a fixed value has been set, the update unit 133 updates the weights of the neural network nodes for which a fixed value has not been set by the setting unit 131 to values within the range between the fixed value and 0.

As shown in equation (3), the update unit 133 can simultaneously optimize MLP51, MLP52, MLP53, and MLP54 using a single error function.

In other words, the update unit 133 updates the weights of the nodes included in the neural networks that have not been set to fixed values by the setting unit 131, so that each of the outputs of the multiple neural networks approaches the associated result data.

The generation unit 134 generates information indicating the causal relationships between multiple feature quantities and the resulting data based on the weights of the nodes in the input layer.

Furthermore, the generation unit 134 generates an adjacency matrix representing the causal relationships between data, or a DAG based on the adjacency matrix.

The generation unit 134 generates an adjacency matrix W(θ) of weights between each node, for example, as shown in equation (4).

In the adjacency matrix W(θ), w _ij is the L2 norm of the weights from the i-th node in the input layer to each node in the first hidden layer of the MLP associated with the causal analysis with data x _j as the result. When i=j, w _ij =0.

For example, _w41 is expressed as in equation (5), and _w24 is expressed as in equation (6).

As mentioned above, a _i,k,j ⁽¹⁾ is the weight from the i-th node in the input layer to the k-th node in the first hidden layer of the MLP associated with the causal analysis with data x _j as the result.

As explained in FIG. 5, when the number of nodes in the first hidden layer is m ₁ and the weight of each node is set to 1/√m ₁ , then w _ij =1.

Furthermore, as explained in FIG. 6, when the number of nodes in the first hidden layer is _m1 and the weight of each node is set to 0, w _ij =0.

Furthermore, as explained in FIG. 7, if the number of nodes in the first hidden layer is m ₁ and the range of the weight of each node is set to [0, 1/√m ₁ ], the range of w _ij is [0, 1].

Furthermore, the generation unit 134 may generate and draw a DAG based on the adjacency matrix W(θ). The DAG generated by the generation unit 134 is called a causal graph. Figure 8 shows an example of a causal graph.

In addition, in the drawn causal graph, the strength of the causality between data (the magnitude of the value of the element in the adjacency matrix) is represented by the thickness of the edge between nodes. Also, if there is no causality between data (if the value of the element in the adjacency matrix is 0), the edge between nodes is not drawn.

[Processing flow of the first embodiment]
The flow of the analysis process by the server 10 will be described with reference to Fig. 9. Fig. 9 is a flowchart showing the flow of the analysis process according to the first embodiment.

As shown in FIG. 9, the server 10 defines a multilayer perceptron MLP(X; θ _j ) (j=1, . . . , d) for each data x _j (step S101).
Number of nodes in the input layer: m ₀ (m ₀ = d)
Number of nodes in the first hidden layer: m ₁
Number of nodes in the output layer: 1
Input: X = (x1, ..., _xd ) (where data _xi is input to the i-th node of the input layer)

For example, Figure 4 shows an MLP with two hidden layers, with five nodes in each of the first and second hidden layers. However, the configuration of the MLP is not limited to that shown in Figure 4. The number of hidden layers of the MLP and the number of nodes in each hidden layer may be any number greater than or equal to one.

Next, the server 10 sets weights from the input layer to the hidden layer of each MLP according to the specified causal relationships (step S102).

Next, the server 10 performs simultaneous optimization under the constraints that guide the DAG so that the output of MLP(X; θ _j ) approaches x _j (step S103).

Then, the server 10 generates an adjacency matrix based on the weights from the input layer of each MLP to the first hidden layer (step S104).

The server 10 outputs the generated adjacency matrix or a DAG based on the adjacency matrix (step S105).

Figure 10 is a flowchart showing the flow of the setting process according to the first embodiment. The process in Figure 10 corresponds to step S102 in Figure 9.

As shown in FIG. 10, first, the server 10 assigns 1 to j (step S201) and starts setting MLP(X; θ _j ) (step S202).

Here, the server 10 assigns 1 to i (step S203).

The server 10 checks the specified causal relationship from the data x _i to the data x _j (step S204).

If the causal relationship from data x _i to data x _j is specified as “causal” (step S204, causal), the server 10 sets the weight from the i-th node in the input layer to the first layer of the hidden layer to 1/√m ₁ (fixed value) (step S205).

Furthermore, if the causal relationship from data x _i to data x _j is specified as “no causality” (step S204, no causality), the server 10 sets the weight from the i-th node in the input layer to the first layer of the hidden layer to 0 (fixed value) (step S206).

When i = j, the server 10 performs the same processing as when the causal relationship is "no causal relationship."

Furthermore, if the causal relationship from data x _i to data x _j is not specified (step S204, causality unknown), the server 10 sets the weight from the i-th node in the input layer to the first hidden layer to a variable value of [0, 1/√m ₁ ] (step S207).

Next, the server 10 increments i by 1 (step S208).

If i = d is not satisfied (step S209, No), the server 10 returns to step S204 and repeats the process.

On the other hand, if i=d (Yes at step S209), the server 10 ends the setting of MLP(X; θ _j ) (step S210).

Then, the server 10 increments j by 1 (step S211).

If j = d is not satisfied (step S212, No), the server 10 returns to step S202 and repeats the process.

On the other hand, if j = d (step S212, Yes), the server 10 terminates the setting process.

[Effects of the First Embodiment]
As described above, the setting unit 131 sets fixed values corresponding to the causal relationships to the weights of nodes in the input layer of the neural network, among the plurality of nodes to which each of the plurality of feature quantities is input, to which feature quantities having a pre-specified causal relationship with the result data are input. The updating unit 133 updates the weights of the nodes included in the neural network that have not been set to fixed values by the setting unit 131, so that the output of the neural network approaches the result data. The generating unit 134 generates information indicating the causal relationships between the plurality of feature quantities and the result data, based on the weights of the nodes in the input layer.

In this way, the server 10 can set fixed values to the neural network weights corresponding to data for which causal relationships are specified, update the weights for which fixed values are not set, and obtain the causal relationships between data based on the updated weights.

The specified causal relationship may be based on the user's on-site knowledge and know-how accumulated over many years, or it may be theoretically guaranteed.

As a result, according to the embodiment, known causal relationships can be appropriately reflected in the analysis results of the causal relationships between feature quantities.

Generally, methods for verifying the truth or validity of the inferred results of causal analysis are not fully established, and it can be difficult to ensure statistical reliability. However, in the embodiments, by performing causal analysis by assuming prior knowledge (specified causal relationships) to be true, it is possible to increase the statistical reliability of the causal analysis by at least the amount of prior knowledge.

In addition, even if fixed values are set for the weights in the input layer, it is believed that the strength of causality and the validity of the model are adjusted in the layers after the intermediate layer. As a result, weights that are not set to fixed values will be updated consistently according to the algorithm.

According to the embodiment, MLP can be considered as an approximation of the nonlinear SEM (Structural Equation Model) described in Non-Patent Document 2, and can be applied to analyses such as intervention effect estimation (Reference: "Estimation of intervention effects using predictive models for time-series data" (https://www.jstage.jst.go.jp/article/pjsai/JSAI2020/0/JSAI2020_1J4GS204/_pdf/-char/ja)).

Furthermore, if the specified prior knowledge is correct, the model can be guided to the true local minimum quickly and accurately.

Furthermore, based on the results of the causal analysis, it becomes possible to reexamine the variables of the imitation learning model. As a result, the embodiment has the effect of improving the performance of the imitation learning model.

[System configuration, etc.]
Furthermore, the components of each device shown in the figure are functional concepts and do not necessarily have to be physically configured as shown. In other words, the specific form of distribution and integration of each device is not limited to that shown, and all or part of the devices can be functionally or physically distributed or integrated in any unit depending on various loads, usage conditions, etc. Furthermore, all or any part of the processing functions performed by each device can be realized by a CPU (Central Processing Unit) and a program analyzed and executed by the CPU, or can be realized as hardware using wired logic. Note that the program may be executed not only by the CPU but also by other processors such as a GPU.

Furthermore, among the processes described in the embodiments, all or part of the processes described as being performed automatically can be performed manually, or all or part of the processes described as being performed manually can be performed automatically using known methods. In addition, the information, including the processing procedures, control procedures, specific names, various data, and parameters shown in the above documents and drawings, can be changed as desired unless otherwise specified.

[program]
In one embodiment, the server 10 can be implemented by installing an analysis program that executes the above-described analysis process as package software or online software on a desired computer. For example, by executing the above-described analysis program on an information processing device, the information processing device can function as the server 10. The information processing device referred to here includes desktop and notebook personal computers. Other information processing devices also include tablet terminals, smartphones, mobile communication terminals such as mobile phones and PHS (Personal Handyphone Systems), and slate terminals such as PDAs (Personal Digital Assistants).

The server 10 can also be implemented as a server that provides services related to the above-mentioned analytical processing to clients, with the terminal device used by the user being the client. For example, the server is implemented as a server device that provides analytical services that take the specification of requested points as input and output analysis results (e.g., an adjacency matrix). In this case, the server may be implemented as a web server, or as a cloud that provides services related to the above-mentioned analytical processing through outsourcing.

Figure 11 shows an example configuration of a computer that executes an analysis program. The computer 1000 includes, for example, memory 1010 and a CPU 1020. The computer 1000 also includes a hard disk drive interface 1030, a disk drive interface 1040, a serial port interface 1050, a video adapter 1060, and a network interface 1070. These components are connected by a bus 1080.

The memory 1010 includes a ROM (Read Only Memory) 1011 and a RAM (Random Access Memory) 1012. The ROM 1011 stores a boot program such as a BIOS (Basic Input Output System). The hard disk drive interface 1030 is connected to a hard disk drive 1090. The disk drive interface 1040 is connected to a disk drive 1100. A removable storage medium such as a magnetic disk or optical disk is inserted into the disk drive 1100. The serial port interface 1050 is connected to a mouse 1110 and a keyboard 1120, for example. The video adapter 1060 is connected to a display 1130, for example.

The hard disk drive 1090 stores, for example, an OS 1091, an application program 1092, a program module 1093, and program data 1094. That is, the programs that define each process of the server 10 are implemented as program modules 1093 in which computer-executable code is written. The program modules 1093 are stored, for example, on the hard disk drive 1090. For example, a program module 1093 for executing processes similar to the functional configuration of the server 10 is stored on the hard disk drive 1090. Note that the hard disk drive 1090 may be replaced by an SSD (Solid State Drive).

Furthermore, the setting data used in the processing of the above-described embodiment is stored as program data 1094, for example, in memory 1010 or hard disk drive 1090. Then, the CPU 1020 reads the program module 1093 or program data 1094 stored in memory 1010 or hard disk drive 1090 into RAM 1012 as needed, and executes the processing of the above-described embodiment.

The program module 1093 and program data 1094 do not necessarily have to be stored on the hard disk drive 1090; they may instead be stored on a removable storage medium and read by the CPU 1020 via the disk drive 1100 or the like. Alternatively, the program module 1093 and program data 1094 may be stored on another computer connected via a network (such as a LAN (Local Area Network) or WAN (Wide Area Network)). The program module 1093 and program data 1094 may then be read by the CPU 1020 from the other computer via the network interface 1070.

REFERENCE SIGNS LIST 1 Plant operation system 10 Server 11 Communication unit 12 Storage unit 13 Control unit 20 Terminal device 30 Plant system 51, 52, 53, 54 MLP
121 History DB
122 Model information 131 Setting unit 132 Calculation unit 133 Update unit 134 Generation unit

Claims (4)

a setting unit that sets a fixed value according to the causal relationship to a weight of a node to which a feature having a pre-specified causal relationship with result data is input among a plurality of nodes in an input layer of the neural network, the node receiving each of a plurality of feature amounts; and sets a zero to a weight of a node to which a feature having a pre-specified no causal relationship with result data is input among the plurality of feature amounts ;
an updating unit that updates, among the weights of the nodes included in the neural network, weights for which a fixed value has not been set by the setting unit to a value included in a range between the fixed value and 0, so that an output of the neural network approaches the result data;
a generation unit that generates information indicating a causal relationship between the plurality of feature amounts and the result data based on weights of the nodes in the input layer;
An analytical device comprising: the setting unit sets a fixed value according to the causal relationship to a weight of a node, which is an input layer node of each of the plurality of neural networks to which each of the plurality of feature amounts is input, and to which a feature amount having a causal relationship with the associated result data is specified in advance, among a plurality of nodes to which each of the plurality of feature amounts is input, for each of the plurality of neural networks to which each of the plurality of feature amounts is associated as the result data;
The analysis device according to claim 1, characterized in that the update unit updates weights of nodes included in the neural networks that have not been set to fixed values by the setting unit so that each output of the plurality of neural networks approaches the associated result data. An analytical method performed by an analytical device, comprising:
a setting step of setting a fixed value according to the causal relationship to a weight of a node to which a feature having a pre-specified causal relationship with result data is input among a plurality of nodes in an input layer of the neural network, the node being input with each of a plurality of feature amounts, and setting a zero to a weight of a node to which a feature having a pre-specified no causal relationship with the result data is input among the plurality of feature amounts ;
an updating step of updating, among the weights of the nodes included in the neural network, weights for which fixed values have not been set in the setting step , to values included in a range between the fixed values and 0 so that the output of the neural network approaches the resultant data;
a generating step of generating information indicating a causal relationship between the plurality of feature quantities and the result data based on weights of the nodes in the input layer;
An analytical method comprising: An analysis program for causing a computer to function as the analysis device according to claim 1 or 2 .