Information processing device, information processing method and program

Abstract

To construct a further appropriate model as a model for performing analysis about a production system or the like.SOLUTION: An information processing device is provided with a processing unit. The processing unit sets a constraint condition including a first range of a parameter value in each of a plurality of parameters included in a first model for performing estimation by inputting a plurality of explanatory variables and corresponding to the plurality of explanatory variables. The processing unit constructs the first model by acquiring a plurality of parameters so as to satisfy the constraint condition.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] An embodiment of the present invention relates to an information processing device, an information processing method, and a program. [Background technology]

[0002] In production systems such as factories (such as semiconductor factories) and plants (such as chemical plants), various types of products are mass-produced. In recent years, it has become possible to acquire large amounts of process data from sensors installed in each manufacturing process at short intervals (for example, daily). Furthermore, by analyzing the accumulated data, it has become possible to implement measures to reduce quality variations. Such measures are intended to improve productivity and yield.

[0003] One such measure is regression analysis, which uses a model (regression model) constructed by machine learning or other methods. A regression model is a model that uses process data such as sensor values, set values, and control values ​​as explanatory variables and quality characteristics as objective variables. Using a regression model makes it possible to analyze the factors (causes) behind variations in quality characteristics. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Patent No. 6208259 [Patent Document 2] Patent Publication No. 2021-174040 Summary of the Invention [Problem to be solved by the invention]

[0005] An object of the present invention is to provide an information processing device, an information processing method, and a program that can construct a more appropriate model as a model for analyzing a production system or the like. [Means for solving the problem]

[0006] According to an embodiment, an information processing apparatus includes a processing unit that sets constraints, including a first range of parameter values, for a plurality of parameters corresponding to the plurality of explanatory variables included in a first model that performs estimation by inputting the plurality of explanatory variables. The processing unit constructs the first model by determining the plurality of parameters that satisfy the constraints. [Brief explanation of the drawings]

[0007] [Figure 1] FIG. 1 is a block diagram of an information processing system according to a first embodiment. [Figure 2] 4 is a flowchart of information processing according to the first embodiment. [Figure 3] FIG. 10 is a block diagram of an information processing system according to a second embodiment. [Figure 4] 10 is a flowchart of information processing according to the second embodiment. [Figure 5] FIG. 10 is a diagram for explaining a method for setting a range of a coefficient. [Figure 6] FIG. 10 is a diagram showing an example of an output display screen. [Figure 7] FIG. 10 is a diagram showing an example of an output display screen. [Figure 8] FIG. 10 is a diagram showing an example of an output display screen. [Figure 9] FIG. 10 is a diagram showing an example of an output display screen. [Figure 10] FIG. 10 is a diagram showing an example of an output display screen. [Figure 11] FIG. 10 is a block diagram of an information processing system according to a third embodiment. [Figure 12] 10 is a flowchart of information processing according to the third embodiment. [Figure 13] FIG. 10 is a block diagram of an information processing system according to a fourth embodiment. [Figure 14] 10 is a flowchart of information processing according to the fourth embodiment. [Figure 15] FIG. 1 is a hardware configuration diagram of an information processing apparatus according to an embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0008] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Preferred embodiments of an information processing apparatus according to the present invention will be described in detail below with reference to the accompanying drawings.

[0009] In factor analysis using regression models, highly interpretable models such as linear models, decision trees, and additive models are often used. For each explanatory variable, a quantity that represents the degree of influence that the model parameters have on the model output is calculated, and by using this calculated quantity, it is possible to identify factors that can explain the variability in quality characteristics.

[0010] The model parameters are, for example, the regression coefficients and importance of a regression model. In the following, an example will be mainly described in which a regression model is used as the model and coefficients (regression coefficients) are used as the model parameters. Applicable models and model parameters are not limited to these.

[0011] Trends in data from production systems and other sources can change from moment to moment. To always grasp the latest trends, it is necessary to regularly update the model using the latest data. On the other hand, if only the most recent data is used, the number of data points will be reduced, which will make the effects of noise more pronounced and may result in significant differences between the estimated coefficients and the experts' perceptions. In manufacturing sites, even the smallest number of coefficients that differ from the perceptions can cause the model itself to lose credibility and may be deemed inappropriate for use.

[0012] To avoid such problems, techniques have been proposed to reconstruct the model after changing the size of the data (sample size) used to estimate the coefficients or changing the random number seed.In addition, when an ideal model in which all coefficients are within the expected range is obtained, techniques have been proposed to build a constrained model that reduces the difference from the coefficients of the ideal model.

[0013] Since production system experts have rich domain knowledge about the relationships between devices, sensors, and quality characteristics, it is rare that the constructed model perfectly matches the domain knowledge.

[0014] Even if technology is applied to reconstruct the model by changing the conditions, there is no guarantee that all coefficients will fall within the expected range determined by the domain knowledge of experts. One possible method is to repeatedly reconstruct the model until the expected results are obtained, but this could increase the processing load.

[0015] Even if constrained model building techniques are applied to minimize the deviation from the ideal model coefficients, there is no guarantee that the coefficients will fall within the expected range. Furthermore, because this technique does not specify a range for the coefficients but sets an ideal value, even a slight deviation from the ideal value increases bias and reduces the accuracy of the model.

[0016] Therefore, the information processing device of the following embodiment imposes a range of coefficients that experts in, for example, production systems consider appropriate as a constraint when constructing a model. This makes it possible to construct a model that includes coefficients within an expected range. For example, the information processing device of the embodiment sets a range of coefficients for each explanatory variable and constructs a model using the coefficient range as a constraint. This makes it possible to obtain a model that includes coefficients within the set range.

[0017] (First embodiment) The following describes an embodiment for constructing a model that can be used for quality control in a production system. As described above, in a production system, measures are taken to improve yield by suppressing variability and fluctuations in quality characteristics and reducing defects. To clarify the causes of variability and fluctuations in quality characteristics, a regression model, for example, is used.

[0018] A product goes through many manufacturing processes before it is finished. When analyzing the causes of variation in the quality characteristics of finished products, a model is constructed using information such as the type of manufacturing equipment used in each manufacturing process and the sensor values ​​detected by sensors installed in the manufacturing equipment as explanatory variables. Information such as the type of manufacturing equipment and sensor values ​​can be interpreted as feature quantities that represent the characteristics of the target of analysis, such as a production system.

[0019] As manufacturing equipment deteriorates over time, the trends in the acquired process data also change. Furthermore, tasks that affect the trends in the process data, such as periodic maintenance and part replacement, may be performed. Therefore, for example, models are updated to reflect changes in the trends in the process data.

[0020] In the model of this embodiment, for example, the dependent variables are quality characteristics, defect rates, and variables indicating whether a product is good or bad. The dependent variables may be sensor values ​​detected by a sensor. The explanatory variables are other sensor values, set values, control values, and the like. Preprocessing may be performed on the explanatory variables in advance. Examples of preprocessing include standardization, scaling, transformation using a specific function, addition of interaction terms, time lag, time lead, conversion to dummy variables, encoding, outlier processing, and missing value processing.

[0021] 1 is a block diagram showing an example of the configuration of an information processing system including an information processing device according to this embodiment. As shown in FIG. 1, the information processing system includes an information processing device 100 and a management system 200 connected via a network 300.

[0022] The information processing device 100 and the management system 200 can each be configured as, for example, a server device. The information processing device 100 and the management system 200 may be realized as multiple physically independent devices (systems), or the respective functions may be configured within a single physical device. In the latter case, the network 300 may not be provided. At least one of the information processing device 100 and the management system 200 may be built on a cloud environment.

[0023] The network 300 is, for example, a network such as a LAN (Local Area Network) or the Internet. The network 300 may be either a wired network or a wireless network. The information processing device 100 and the management system 200 may transmit and receive data using a direct wired or wireless connection between components without going through the network 300.

[0024] The management system 200 is a system that manages data used for model learning (construction, updating), analysis, etc. The management system 200 includes a storage unit 221 and a communication control unit 201.

[0025] The storage unit 221 stores various types of information used in various processes executed by the management system 200. For example, the storage unit 221 stores data (process data, etc.) including a target variable and an explanatory variable. The storage unit 221 can be configured using any commonly used storage medium, such as a flash memory, a memory card, a RAM (Random Access Memory), an HDD (Hard Disk Drive), or an optical disk.

[0026] The communication control unit 201 controls communication with an external device such as the information processing device 100. For example, the communication control unit 201 transmits process data to the information processing device 100.

[0027] Each of the above units (communication control unit 201) is realized, for example, by one or more processors. For example, each of the above units may be realized by causing a processor such as a CPU (Central Processing Unit) or a GPU (Graphics Processing Unit) to execute a program, i.e., by software. Each of the above units may be realized by a processor such as a dedicated IC (Integrated Circuit), i.e., by hardware. Each of the above units may be realized by using a combination of software and hardware.

[0028] The information processing device 100 includes a target data storage unit 121, an input device 131, a display 132, a communication control unit 101, an acquisition unit 102, a setting unit 103, a construction unit 104, an accuracy calculation unit 105, and an output control unit 106.

[0029] The target data storage unit 121 stores various information used in various processes executed by the information processing device 100. For example, the target data storage unit 121 stores information (such as process data) acquired from the management system 200 via the communication control unit 101 and the acquisition unit 102, and parameters (coefficients) of the model constructed by the construction unit 104. The target data storage unit 121 can be configured using any commonly used storage medium, such as a flash memory, a memory card, a RAM, a HDD, or an optical disc.

[0030] The input device 131 is a device for a user or the like to input information. The input device 131 is, for example, a keyboard and a mouse. The display 132 is an example of an output device that outputs information, and is, for example, a liquid crystal display. The input device 131 and the display 132 may be integrated into one device, such as a touch panel.

[0031] The communication control unit 101 controls communication with external devices such as the management system 200. For example, the communication control unit 101 receives process data and the like from the management system 200. The communication control unit 101 also transmits to the management system 200 a request to transmit process data for a specified period.

[0032] The acquiring unit 102 acquires various types of information. For example, the acquiring unit 102 acquires process data received from the management system 200 via the communication control unit 201 and the communication control unit 101.

[0033] For example, the acquiring unit 102 acquires process data for a specified period or process data of a specified sample size from the management system 200 via the communication control unit 101 as data to be analyzed (target data).

[0034] The setting unit 103 sets constraint conditions used when the construction unit 104 constructs the model MA. The model MA is a model (first model) that estimates a response variable by inputting a plurality of explanatory variables. The constraint conditions are constraint conditions on a plurality of coefficients (parameters) included in the model MA. For example, the constraint conditions include a range R1 (first range) of coefficient values ​​for each of the plurality of coefficients.

[0035] The constructing unit 104 constructs the model MA by determining a plurality of coefficients that satisfy the constraints set by the setting unit 103 .

[0036] The accuracy calculation unit 105 calculates the accuracy of the constructed model MA. For example, the accuracy calculation unit 105 performs estimation using the model MA using the process data used for construction or process data different from the process data used for construction as verification data, and calculates the accuracy of the model MA based on the difference between the estimated objective variable and the objective variable included in the process data.

[0037] The output control unit 106 controls the output of various information used in the information processing device 100. For example, the output control unit 106 displays information (such as accuracy and coefficients) of the model MA constructed by the construction unit 104 on the display 132. This allows, for example, an expert to determine whether the estimated coefficients are within expected ranges.

[0038] At least a part of each of the above units (communication control unit 101, acquisition unit 102, setting unit 103, construction unit 104, accuracy calculation unit 105, and output control unit 106) may be realized by one or more processing units. Each of the above units is realized, for example, by one or more processors. For example, each of the above units may be realized by having a processor such as a CPU or GPU execute a program, that is, by software. Each of the above units may be realized by a processor such as a dedicated IC, that is, by hardware. Each of the above units may be realized by using a combination of software and hardware. When multiple processors are used, each processor may realize one of the units, or may realize two or more of the units.

[0039] Next, the details of the processing by each unit of the information processing device 100 will be further explained.

[0040] In the following, it is assumed that the data (target data, process data) acquired by the acquisition unit 102 is n in total (n is an integer of 2 or more), and each data contains numerical values ​​representing p (p is an integer of 1 or more) explanatory variables and one objective variable. That is, the data is (x i ,y i ),x i ∈R p ,y i ∈R,i=1,···,n. x i is the explanatory variable of the p-dimensional vertical vector. y i is a scalar response variable.

[0041] For example, in the case of a linear regression model, the constructing unit 104 estimates the parameters of the model MA, that is, the coefficients for each explanatory variable, by using the acquired data to solve the optimization problem expressed by the following equation (1).

number

[0042] The vector β is the explanatory variable x i Coefficient β set for each p element of j (j is an integer satisfying 1≦j≦p) is a vector whose elements are β j_low and β j_up are the coefficients β j The lower and upper limits of the value of β j_low ≦β j ≦β j_up " is the coefficient β j That is, the setting unit 103 sets the lower limit β as shown in the formula (1), for example. j_low and upper limit β j_up A constraint condition representing the range R1 is set by:

[0043] Explanatory variable x i In order to configure the constraint conditions for the coefficients of each of the p elements of the variable x, the setting unit 103 acquires information on whether or not there is a constraint from the user or the management system 200. The information on whether or not there is a constraint is obtained by i For coefficients for which the constraint is "no", the setting unit 103 does not impose a constraint or imposes a constraint by β j_low =-∞, β j_up =+∞ is added as a range R1 (constrained range). For a constrained coefficient, the setting unit 103 sets at least one of an upper limit and a lower limit as the range R1 of the coefficient. If the unset side is the upper limit, +∞ may be set, and if it is the lower limit, -∞ may be set. Also, if the upper limit is a positive value and no lower limit is set, the lower limit may be set to zero. If the lower limit is a negative value and no upper limit is set, the upper limit may be set to zero. The setting unit 103 may set both a lower limit and an upper limit as the range R1 of the coefficient.

[0044] The term Σ in equation (1) corresponds to a loss function. The construction unit 104 uses such a loss function to find a plurality of parameters (coefficients) that minimize the loss calculated by the loss function, thereby constructing the model MA. That is, by using equation (1), the construction unit 104 finds the coefficient β j We can estimate β̂ under constraints that represent the range R1 of values ​​of

[0045] Although equation (1) uses a loss function using squared error, the loss function is not limited to this. For example, the loss function may be absolute value loss, quantile loss, Huber loss, cross entropy loss, epsilon sensitivity loss, logistic loss, 0-1 loss, exponential loss, hinge loss, or smoothed hinge loss. Furthermore, a loss function weighted according to the reliability and date and time of each data may also be used.

[0046] The loss function may include penalties (regularization terms) such as Ridge, Lasso (Least absolute shrinkage and selection operator), SCAD (Smoothly Clipped Absolute Derivation), MCP (Minimax Concave Penalty), Lq norm (0≦q<1), and Elastic Net, in addition to penalties related to coefficient range constraints.

[0047] Furthermore, the model to which this embodiment can be applied is not limited to a linear regression model, and any model that is expressed using parameters may be used, such as a logistic regression model, a Poisson regression model, a generalized linear model, a generalized additive model, a decision tree, or a neural network.

[0048] The loss function including the constraint on the coefficient range is not limited to equation (1) and may be any other function. For example, the coefficient range may be constrained using a link function, as in the following equation (2).

number

[0049] In equation (2), α and x i are α j and x ij In addition, when a logistic function is used as the link function η(·), the coefficient β can be expressed as in equation (3) below.

number

[0050] The link function is not limited to the logistic function, and for example, an exponential function, a logarithmic function, a softmax function, or the like may be used.

[0051] The construction unit 104 may use any optimization method to solve the optimization problem, but for example, the following method can be applied. (M1) Use an optimization solver to perform linear or nonlinear sequential optimization. (M2) The algorithms used in the optimization solver include the interior point method and the trust region reflective method. (M3) Repeat steps 3-1 to 3-4 below until all coefficients satisfy the constraints. 3-1: Perform unconstrained parameter estimation. 3-2: Explanatory variables whose coefficients are outside the R1 range (explanatory variables x i p elements of the explanatory variable x whose coefficient is the furthest or closest to the upper or lower limit of the range R1 k (k is an integer satisfying 1≦k≦p). 3-3: Selected explanatory variable x k Coefficient β of k Set to the upper or lower bound, whichever is closer. 3-4: Let the objective variable be y i_new =y i -x k β k and the explanatory variable x kExclude. (M4) Allow coefficient values ​​to exceed a given range.

[0052] We will now explain the details of (M4). The above loss function can be interpreted as an example that uses a constraint (hard constraint) that indicates that the coefficients must fall within a given range. The loss function can be configured to use a constraint (soft constraint) that allows the coefficient values ​​to exceed the given range, while keeping them within the range as much as possible. (M4) corresponds to an optimization method that uses such a loss function. The following equation (4) shows an example of a loss function that can be applied to (M4). Equation (4) can be interpreted as a loss function in which the loss becomes smaller as the constraint conditions are satisfied.

number

[0053] λ is a parameter that adjusts the balance between the square sum error and the penalty term. The function L(·) can be the Lq(0<=q) norm, an exponential function, a logarithmic function, etc. The function g(·) can also use the reference value β j_ref is given as a parameter, and if "else", then |β j -β j_ref | may be returned as the result of g(·).

[0054] When the construction of the model is completed, a coefficient for each explanatory variable is obtained. The accuracy calculation unit 105 can calculate the accuracy of the constructed model. The output control unit 106 may present the calculated accuracy of the model directly to the user via an output device such as the display 132. When a reference value of the model accuracy has been obtained, the output control unit 106 may output both the reference value and the calculated accuracy. The output control unit 106 may also output both the coefficient for each explanatory variable and at least one of an upper limit and a lower limit set for the coefficient.

[0055] The lower and upper limit values ​​of the range R1 set by the setting unit 103 are specified, for example, by a user (expert) or the management system 200. Different values ​​may be set for the lower and upper limits according to two or more levels.

[0056] Below, we will explain an example of switching the lower and upper limits according to multiple levels. Considering the use in quality control in mass production of products, the explanatory variable (explanatory variable x i A different coefficient control level may be set for each of the p elements of the product. The method for setting the range R1 differs for each control level. Below, we will explain how to control the coefficient range of each explanatory variable at three control levels, using the finished concentration of a certain material contained in a manufactured product as the target variable. (High-level) If the injection concentration of a material used in the manufacture of a product is an explanatory variable, the relationship between the explanatory variable and the target variable can be determined from domain knowledge, and relatively narrow upper and lower limits of the coefficient are set based on domain knowledge. (Medium level) When the temperature of the manufacturing environment is an explanatory variable, it is clear that there is a positive relationship with the objective variable, so the upper limit of the coefficient range is set to ∞ and the lower limit to 0. In other words, only the sign of the coefficient is managed. (Low-level) For explanatory variables whose relationship with the objective variable is unclear, the upper limit is set to ∞ and the lower limit is set to -∞. In this case, the explanatory variables may be excluded from the management targets and configured not to be included in the constraints when building the model.

[0057] In this way, by setting an appropriate coefficient range R1 for each management level of the explanatory variables and building a model using the set range R1 as a constraint, it is possible to guarantee the construction of a model with coefficients that fall within the range expected by the manufacturing site expert. In addition, by simultaneously visualizing and presenting the estimated coefficients and their ranges to the expert, it is possible to improve the expert's confidence in the model.

[0058] Next, a description will be given of information processing by the information processing apparatus 100 of the first embodiment. Fig. 2 is a flowchart showing an example of information processing in the first embodiment.

[0059] The acquiring unit 102 acquires target data for analysis (step S101). For example, the acquiring unit 102 transmits a transmission request for process data for a period or a size designated by a user or the like to the management system 200, and acquires the process data transmitted from the management system 200 in response to the transmission request as target data.

[0060] The setting unit 103 sets a range R1 of coefficients as a constraint used when constructing the model MA using the acquired target data (process data) (step S102). The constructing unit 104 constructs the model MA by obtaining a plurality of coefficients that satisfy the set constraint (step S103).

[0061] The accuracy calculation unit 105 calculates the accuracy of the constructed model MA (step S104). The output control unit 106 outputs the coefficients of the constructed model MA and the calculated accuracy (step S105), and the information processing ends.

[0062] In this way, the information processing device of the first embodiment can build a model that includes coefficients within the expected range by applying the range of coefficients specified by an expert or the like as a constraint when building a model, thereby enabling the building of a more appropriate model.

[0063] (Second embodiment) In the first embodiment, the range R1 for each coefficient is specified by, for example, a user when constructing a model. In the second embodiment, a more plausible coefficient range R1 is set based on at least one of domain knowledge and coefficients of models constructed in the past.

[0064] For example, in production systems, models are updated periodically, and information on previously constructed models is often retained for purposes such as analyzing abnormal factors. In such cases, the trend in the coefficients of specific explanatory variables over past normal periods can serve as a basis for determining the appropriate range of the coefficients. Meanwhile, the manufacturing process of semiconductors, for example, involves many steps. The vast amount of domain knowledge related to such manufacturing processes is managed in the form of databases and files.

[0065] Therefore, the information processing device of this embodiment determines a reasonable range of coefficients and sets the range of coefficients by referring to information about past models (model history) and information representing domain knowledge, without requiring information input from the user.

[0066] Fig. 3 is a block diagram showing an example of the configuration of an information processing system including an information processing device 100-2 according to the second embodiment. As shown in Fig. 3, the information processing system is configured such that the information processing device 100-2 and a management system 200 are connected via a network 300. The management system 200 and the network 300 are the same as those in the first embodiment, and therefore are denoted by the same reference numerals and will not be described here.

[0067] The information processing device 100-2 includes a target data storage unit 121, a model data storage unit 122-2, a knowledge data storage unit 123-2, an input device 131, a display 132, a communication control unit 101, an acquisition unit 102, a setting unit 103-2, a construction unit 104-2, an accuracy calculation unit 105, an output control unit 106-2, and a range calculation unit 107-2.

[0068] The second embodiment differs from the first embodiment in that a model data storage unit 122-2, a knowledge data storage unit 123-2, and a range calculation unit 107-2 are added, and in the functions of a setting unit 103-2, a construction unit 104-2, and an output control unit 106-2. The other configurations and functions are the same as those of the information processing device 100 of the first embodiment in FIG. 1, which is a block diagram of the information processing device 100, and therefore the same reference numerals are used and the description thereof will be omitted here.

[0069] The model data storage unit 122-2 stores model data relating to one or more models (past models) constructed in the past. The past models are one or more models (second models) already constructed to which a plurality of explanatory variables similar to the model MA to be constructed are input. Hereinafter, the past models may be referred to as models MB.

[0070] The model data includes, for example, the names of explanatory variables (explanatory variable names) and upper and lower limits of coefficients corresponding to the explanatory variables. The model data of the model MB may be acquired from the management system 200 or the like and stored in the model data storage unit 122-2. The model MB may be a model previously constructed by the information processing device 100, or may be a model previously constructed by another device such as the management system 200.

[0071] The knowledge data storage unit 123-2 stores knowledge data representing domain knowledge obtained in advance by, for example, an expert, etc. The knowledge data does not need to be set for all explanatory variables (coefficients), but is set in advance for at least some of the multiple explanatory variables (coefficients).

[0072] The knowledge data includes, for example, the name of the explanatory variable, the presence or absence of a constraint, an upper limit, and a lower limit. The presence or absence of a constraint is information indicating whether or not a constraint is imposed on the coefficient of the corresponding explanatory variable. For example, the presence or absence of a constraint is marked with "o" if a constraint is imposed, and marked with "x" if no constraint is imposed. The knowledge data may be acquired from the management system 200 or the like, and stored in the knowledge data storage unit 123-2.

[0073] The range calculation unit 107-2 calculates the upper and lower limits of the corresponding coefficients for each explanatory variable using the model data of the model MB. For example, the range calculation unit 107-2 calculates the upper and lower limits for each of the multiple coefficients included in the model MB using a range R2 (second range) of coefficient values. The range R2 is, for example, the range from the minimum value to the maximum value of the coefficient values ​​of each of one or more model MBs.

[0074] The range calculation unit 107-2 calculates, for example, the value corresponding to the a percentile (the first percentile) determined by the specified value a among the values in the range R2 as the upper limit. Further, the range calculation unit 107-2 calculates, as the lower limit, the value corresponding to the b percentile (the second percentile) determined by the specified value b (b is a real number satisfying 0 < b < a) among the values in the range R2. The values a and b are specified, for example, by the user.

[0075] The setting unit 103-2 sets the constraint conditions used when constructing the model MA using at least one of the knowledge data and the model data. When using the knowledge data, the setting unit 103-2 sets the constraint conditions (range R1) using, for example, the knowledge data representing the lower limit L1 (the first lower limit) and the upper limit U1 (the first upper limit) of the coefficient values.

[0076] When using the model data, the setting unit 103-2 sets the constraint conditions (range R1) using the range R₂. For example, the setting unit 103-2 sets the range from the lower limit to the upper limit calculated by the range calculation unit 107-2 using the range R₂ as the range R1. Hereinafter, the lower limit and the upper limit calculated by the range calculation unit 107-2 using the range R₂ are represented as the lower limit L₂ and the upper limit U₂, respectively. [[ID=eleven]]

[0077] When using both the knowledge data and the model data, the setting unit 103-2 sets the constraint conditions (range R1) using the statistical values (such as the maximum value, the minimum value, the average value, etc.) of the upper limit or the lower limit obtained from the knowledge data and the upper limit or the lower limit obtained from the model data. For example, the setting unit 103-2 sets the range with the statistical value of the upper limit U1 obtained from the knowledge data and the upper limit U2 obtained from the model data as the upper limit, and the statistical value of the lower limit L1 obtained from the knowledge data and the lower limit L2 obtained from the model data as the lower limit as the range R1.

[0078] The construction unit 104-2 is similar to the construction unit 104 of the first embodiment in that it constructs a model MA by calculating a plurality of coefficients that satisfy the constraint conditions set by the setting unit 103-2. The construction unit 104-2 may further include a function for constructing a model without constraint conditions. For example, the construction unit 104-2 calculates an unconstrained model, which is a model constructed by calculating a plurality of coefficients without using constraint conditions. The unconstrained model can be used, for example, for comparison with the model MA constructed with constraint conditions.

[0079] The output control unit 106-2 differs from the output control unit 106 of the first embodiment in that it further outputs information related to the model MB. For example, the output control unit 106-2 outputs a plurality of coefficients included in the model MA and a plurality of coefficients included in the model MB together. When an unconstrained model is obtained, the output control unit 106-2 may output a plurality of coefficients included in the model MA constructed under constraint conditions together with a plurality of coefficients included in the unconstrained model.

[0080] Next, information processing by the information processing device 100-2 of the second embodiment will be described with reference to Fig. 4. Fig. 4 is a flowchart showing an example of information processing in the second embodiment.

[0081] Step S201 is the same process as step S101 in the information processing device 100 of the first embodiment, and therefore a description thereof will be omitted.

[0082] The range calculation unit 107-2 reads out model data of the model MB (past model) from, for example, the model data storage unit 122-2, and calculates the range of the coefficients (lower limit L2, upper limit U2) using the read out model data (step S202).

[0083] The setting unit 103-2 sets a coefficient range R1 as a constraint condition used when constructing the model MA, based on the calculated range (the range from the lower limit L2 to the upper limit U2) and the knowledge data read from the knowledge data storage unit 123-2 (step S203).

[0084] Steps S204 to S206 are the same as steps S103 to S105 in the information processing apparatus 100 of the first embodiment, and therefore a description thereof will be omitted.

[0085] Next, the method for setting the ranges of the coefficients in this embodiment will be described in further detail below with reference to Fig. 5, which is a diagram for explaining the method for setting the ranges of the coefficients.

[0086] The domain knowledge in Fig. 5 corresponds to the knowledge data stored in the knowledge data storage unit 123-2, for example. For example, as domain knowledge related to "valve flow rate", knowledge data is obtained in which the presence or absence of constraint is "○" (constrained), the upper limit of the corresponding coefficient is 4, and the lower limit of the coefficient is 1. In Fig. 5, the upper and lower limits obtained as domain knowledge are the upper limit and the lower limit, respectively. domain and lower bound domain It is expressed as follows. domain and lower bound domain correspond to the upper limit U1 and lower limit L1 above, respectively.

[0087] The graph on the right of Figure 5 is a graph showing the change in the coefficient corresponding to a certain explanatory variable (for example, "valve flow rate"). Such changes in the coefficient can be obtained from the model data of multiple models MB (past models). In other words, it is possible to obtain the change in the value of the coefficient of the corresponding explanatory variable from each of the models MB constructed at multiple times in the past.

[0088] The central table in FIG. 5 shows an example of the range of coefficients calculated from the past model. For example, the central table shows an example of the lower and upper limits calculated by the range calculation unit 107-2 using the range R2 determined by the minimum and maximum values ​​of the coefficients as shown in the graph on the right. As described above, the upper and lower limits are values ​​within the range R2 corresponding to the a percentile and the b percentile, respectively. In FIG. 5, the upper and lower limits obtained from the past model MB are the upper and lower limits, respectively. past and lower bound past It is expressed as follows. past and lower boundpast correspond to the upper limit U2 and lower limit L2 above, respectively.

[0089] 5, for explanatory variables with constraints, the upper and lower limits of the coefficients are obtained from both the domain knowledge and the history of the model MB. For example, the setting unit 103-2 uses these to determine the upper and lower limits that represent the final range of the coefficients. final and lower bound final Set as follows: ·upper limit final is the upper limit domain and upper limit past The maximum value among these is used. ·lower limit final is the lower bound domain and lower bound past The minimum value among these.

[0090] The method for setting the final range is not limited to the above. For example, instead of the maximum value, an upper limit domain and upper limit past Other statistical values ​​(minimum, average, etc.) final Alternatively, instead of the minimum value, a lower limit may be set. domain and lower bound past Other statistical values ​​(max, average, etc.), lower bound final may be set.

[0091] The setting unit 103-2 sets the upper limit domain and upper limit past Either of the above, final Similarly, the setting unit 103-2 may set the lower limit domain and lower bound past Either one of the two is the lower limit final Which of the domain knowledge and the history of the model MB is to be given priority may be determined according to information stored in advance in a storage device in the information processing device 100, for example.

[0092] The setting unit 103-2 sets the range of coefficients obtained from the history of the model MB to the range of coefficients obtained from the domain knowledge (lower limit domain to upper limitdomain The range that is most similar to the final range (the range up to the lower limit final to upper limit final The range may be set as

[0093] In addition, if there is no overlapping range between the range of coefficients obtained from domain knowledge and the range of coefficients obtained from the history of the model MB, a loss function that imposes a penalty (e.g., an L1 penalty) on the corresponding coefficients may be used.

[0094] The construction unit 104-2 constructs the model MA using the final coefficient range R1 set by the setting unit 103-2. The construction unit 104-2 may use any of the optimization methods described above. The construction unit 104-2 may determine the optimization method to be applied in accordance with, for example, a user's specification.

[0095] There are cases where it is difficult to determine the appropriateness of the upper and lower limit values ​​based solely on the domain knowledge of the expert, or where it is not possible to respond to sudden changes because the physical relationships are not known based solely on the history of the model MA. Setting the coefficient range by referring to both the domain knowledge and the history of the model MB can avoid these problems and set a more appropriate range R1. Furthermore, for example, experts only need to update their domain knowledge when necessary, reducing the processing load required for updating.

[0096] Note that, for each explanatory variable, if only one of the domain knowledge and the history of the model MB is available, the range of the corresponding coefficient may be set using only the available information.

[0097] Furthermore, the setting unit 103-2 may set the range of the coefficients according to the distribution of the signs of the values ​​of the multiple coefficients included in the model MB. For example, when the proportion of the corresponding coefficients having a certain sign (positive or negative) for a certain explanatory variable is equal to or greater than a specified threshold, the setting unit 103-2 may set the range of the coefficients (lower limit) so that the range corresponds to the sign. past ,upper limit past ) may be set. The range according to the sign is, for example, if the sign is positive, the lower limit is past =0, upper limit past =∞, and if the sign is negative, the lower bound past =-∞, upper limit past =0.

[0098] Next, an example of output by the output control unit 106-2 will be described. As described above, the output control unit 106-2 may output the coefficients from the model MB (past model) together with the estimated coefficients and the coefficient ranges determined by the system. The output control unit 106-2 may output the values ​​of the coefficients, or may output a graph representing the coefficients instead of the coefficient values.

[0099] 6 to 10 are diagrams showing examples of display screens output (displayed) by the output control unit 106-2. Each diagram corresponds to an example of a display screen output for one explanatory variable F1. The circles in Fig. 6 and Figs. 8 to 10 represent estimated coefficients. In Fig. 7, the shaded areas correspond to estimated coefficients.

[0100] Figure 6 is an example of a display screen represented by a box plot. Figure 7 is an example of a display screen represented by a bar graph. Figure 8 is an example of a display screen represented by a scatter plot. Figure 9 is an example of a display screen represented by a distribution fitted using a non-parametric method. Figure 10 is an example of a display screen including lines representing lower and upper limits of the coefficients and estimated coefficients.

[0101] In this way, in the second embodiment, it is possible to set a more reasonable range of coefficients based on at least one of domain knowledge and coefficients of models constructed in the past.

[0102] (Third embodiment) In the third embodiment, similar to the second embodiment, a more appropriate coefficient range R1 is set using domain knowledge and information on the model MB (past model) to construct the model MA. In the second embodiment, the coefficient range R1 determined from the domain knowledge and information on the model MB is used as is to construct the model MA. In this method, the function for setting the coefficient range becomes a black box for the user, and the user cannot know in detail how the range was set.

[0103] Therefore, the information processing device of the third embodiment presents the user with upper and lower limits determined from domain knowledge and the history of the model MB, and sets the upper and lower limit values ​​specified (adjusted) by the user as the final range R1 by referring to the presented information.

[0104] Fig. 11 is a block diagram showing an example of the configuration of an information processing system including an information processing device 100-3 according to the third embodiment. As shown in Fig. 11, the information processing system has a configuration in which the information processing device 100-3 and a management system 200 are connected via a network 300. The management system 200 and the network 300 are the same as those in the first embodiment, and therefore are denoted by the same reference numerals and will not be described here.

[0105] The information processing device 100-3 includes a target data storage unit 121, a model data storage unit 122-2, a knowledge data storage unit 123-2, an input device 131, a display 132, a communication control unit 101, an acquisition unit 102, a setting unit 103-3, a construction unit 104-2, an accuracy calculation unit 105, an output control unit 106-3, and a range calculation unit 107-2.

[0106] In the third embodiment, the functions of the setting unit 103-3 and the output control unit 106-3 are different from those in the second embodiment. The other configurations and functions are the same as those in the block diagram of the information processing device 100 in the second embodiment shown in FIG. 3, and therefore the same reference numerals are used and the description thereof will be omitted here.

[0107] The setting unit 103-3 is similar to the setting unit 103-2 in the second embodiment in that it uses at least one of knowledge data and model data to determine the lower and upper limits of the range of the coefficients.

[0108] The output control unit 106-3 differs from the output control unit 106-2 of the second embodiment in that it further includes a function for outputting the lower and upper limits of the range of coefficients determined by the setting unit 103-3. For example, the output control unit 106-3 outputs output information including at least one of knowledge data representing the lower limit L1 and the upper limit U1, and the lower limit L2 and the upper limit U2 determined from the range R2 of coefficients included in the model MB.

[0109] In this embodiment, the setting unit 103-3 differs from the setting unit 103-2 of the second embodiment in that it sets constraint conditions (range R1) specified (adjusted) by, for example, a user based on the output information output by the output control unit 106-3.

[0110] Next, information processing by the information processing device 100-3 of the third embodiment will be described with reference to Fig. 12. Fig. 12 is a flowchart showing an example of information processing in the third embodiment.

[0111] Steps S301 to S302 are the same as steps S201 to S202 in the information processing device 100-2 of the second embodiment, and therefore a description thereof will be omitted.

[0112] The output control unit 106-3 displays output information including the coefficient range (lower limit L2, upper limit U2) set in step S302 and the coefficient range (lower limit L1, upper limit U1) obtained from the knowledge data, for example, on the display 132 (step S303).

[0113] The setting unit 103-3 refers to the displayed output information and sets, for example, the range of coefficients designated by the user as the range R1 of coefficients as a constraint used when constructing the model MA (step S304).

[0114] Steps S305 to S307 are the same as steps S204 to S206 in the information processing device 100-2 of the second embodiment, and therefore description thereof will be omitted.

[0115] Next, an example of a method for displaying output information will be described. For example, the output control unit 106-3 displays output information in which, for each explanatory variable, the upper limit U2 and lower limit L2 obtained from the history of the model MB, and the upper limit U1 and lower limit L1 obtained from the domain knowledge are displayed together as a graph or text.

[0116] The output control unit 106-3 may output the output information so that the output modes of the lower limit and the upper limit are different from each other. The output modes include, but are not limited to, color and shape, for example. The output control unit 106-3 may output the output information represented by a scroll bar corresponding to the lower limit L1 to the upper limit U1 and a scroll bar corresponding to the lower limit L2 to the upper limit U2. The output control unit 106-3 may obtain a distribution of coefficient values ​​from the history of the model MB and output the obtained distribution as the output information.

[0117] The information processing device 100-3 may further include a function of updating the domain knowledge using the range adjusted by the user.

[0118] In this way, in the third embodiment, the range of coefficients (upper and lower limits) determined by the information processing device is output, and the user can adjust the range of coefficients. That is, the user can determine the final range R1.

[0119] (Fourth embodiment) In the second embodiment, a more appropriate coefficient range is set by utilizing the coefficients of the model MB (past model). In the fourth embodiment, information on a model in a similar environment is utilized instead of the model MB.

[0120] Fig. 13 is a block diagram showing an example of the configuration of an information processing system including an information processing device 100-4 according to the fourth embodiment. As shown in Fig. 13, the information processing system has a configuration in which the information processing device 100-4 and a management system 200 are connected via a network 300. The management system 200 and the network 300 are the same as those in the first embodiment, and therefore are denoted by the same reference numerals and will not be described here.

[0121] The information processing device 100-4 includes a target data storage unit 121, a model data storage unit 122-4, an input device 131, a display 132, a communication control unit 101, an acquisition unit 102, a setting unit 103-4, a construction unit 104-2, an accuracy calculation unit 105, an output control unit 106-3, and a range calculation unit 107-4.

[0122] The third embodiment differs from the second embodiment in that the knowledge data storage unit 123-2 is deleted and in that the functions of the model data storage unit 122-4, the setting unit 103-4, and the range calculation unit 107-4 are different. The other configurations and functions are the same as those of the information processing device 100 of the second embodiment in FIG. 3, which is a block diagram of the information processing device 100, and therefore the same reference numerals are used and the description thereof will be omitted here.

[0123] The model data storage unit 122-4 stores model data relating to one or more models (similar models) similar to the model MA to be constructed. A similar model is a model (third model) that performs estimation by inputting at least some of the multiple explanatory variables of the model MA to be constructed. Hereinafter, the similar model may be referred to as a model MC. Model data for the model MC may be acquired from the management system 200 or the like and stored in the model data storage unit 122-4.

[0124] The similar model (model MC) corresponds to a model used in an environment similar to the environment in which the model MA is used, for example. The similar environment is, for example, the following multiple environments. Multiple environments for multiple machines produced by the same manufacturer Multiple environments for multiple product varieties produced in the same factory Multiple environments using the same type of sensors installed in different locations Multiple environments each testing the same product

[0125] The range calculation unit 107-4 calculates the upper and lower limits of the corresponding coefficients for each explanatory variable using the model data of the model MC. The function of the range calculation unit 107-4 is the same as that of the range calculation unit 107-2 in the second embodiment, except that the model MC is used instead of the model MB.

[0126] For example, the range calculation unit 107-4 calculates upper and lower limits for each of the coefficients included in the model MC that have the same or similar names or meanings as those of the model MA, using a range R3 of coefficient values. The range R3 is, for example, the range from the minimum value to the maximum value of the coefficients of one or more models MC. The range calculation unit 107-4 can identify coefficients with the same or similar names or meanings, for example, using correspondence information that indicates the correspondence between predetermined coefficients of the model MA and the coefficients of the model MC.

[0127] For example, range calculation unit 107-4 calculates the upper limit of the range R3 to be the value corresponding to the a percentile determined by the specified value a, and the lower limit of the range R3 to be the value corresponding to the b percentile determined by the specified value b.

[0128] The setting unit 103-4 uses the range R3 to set the constraint condition (range R1) used when constructing the model MA. For example, the setting unit 103-4 sets the range from the lower limit to the upper limit calculated by the range calculation unit 107-3 using the range R3 as the range R1. Hereinafter, the lower limit and upper limit calculated by the range calculation unit 107-4 using the range R3 will be referred to as the lower limit L3 and the upper limit U3, respectively.

[0129] There may be multiple similar environments similar to the environment in which the model MA is used. Furthermore, multiple models MC constructed at multiple times in the past may be used in each similar environment. That is, for each explanatory variable, the upper limit U3 and the lower limit L3 may be determined using up to l × m coefficients obtained from m (m is an integer equal to or greater than 1) models MC in l (l is an integer equal to or greater than 1) similar environments.

[0130] In addition, the second embodiment and the fourth embodiment may be combined to determine the final upper and lower limits (range R1) using one or more of the upper and lower limits determined from the history of model MC, the upper and lower limits determined from domain knowledge, and the upper and lower limits determined from model MB.

[0131] Next, information processing by the information processing device 100-4 of the fourth embodiment will be described with reference to Fig. 14. Fig. 14 is a flowchart showing an example of information processing in the fourth embodiment.

[0132] Step S401 is the same process as step S201 in the information processing device 100-2 of the second embodiment, and therefore a description thereof will be omitted.

[0133] The range calculation unit 107-4 reads out the model data of the model MC from, for example, the model data storage unit 122-4, and calculates the range of the coefficients (lower limit L3, upper limit U3) using the read out model data (step S402).

[0134] The setting unit 103-4 sets a range R1 of coefficients as a constraint condition used when constructing the model MA from the calculated range (range from the lower limit L3 to the upper limit U3) (step S403).

[0135] Steps S404 to S406 are the same as steps S204 to S206 in the information processing device 100-2 of the second embodiment, and therefore description thereof will be omitted.

[0136] In this way, in the fourth embodiment, even if the history of domain knowledge and model MB (past model) cannot be obtained, or the amount of history of domain knowledge and model MB is small, by referring to information on models of similar environments, a range of reasonable coefficients can be determined and used as a constraint when constructing a model.

[0137] As described above, according to the first to fourth embodiments, a more appropriate model can be constructed as a model for analyzing a production system or the like.

[0138] Next, the hardware configuration of the information processing apparatus according to the first to fourth embodiments will be described with reference to Fig. 15. Fig. 15 is an explanatory diagram showing an example of the hardware configuration of the information processing apparatus according to the first to fourth embodiments.

[0139] The information processing device of the first to fourth embodiments includes a control device such as a CPU (Central Processing Unit) 51, a storage device such as a ROM (Read Only Memory) 52 and a RAM (Random Access Memory) 53, a communication I / F 54 that connects to a network and communicates, and a bus 61 that connects each part.

[0140] The programs executed by the information processing apparatuses of the first to fourth embodiments are provided in advance in the ROM 52 or the like.

[0141] The programs executed by the information processing devices of the first to fourth embodiments may be configured to be provided as a computer program product by being recorded in an installable or executable file format on a computer-readable recording medium such as a CD-ROM (Compact Disk Read Only Memory), a flexible disk (FD), a CD-R (Compact Disk Recordable), or a DVD (Digital Versatile Disk).

[0142] Furthermore, the programs executed by the information processing apparatuses of the first to fourth embodiments may be stored on a computer connected to a network such as the Internet and provided by being downloaded via the network. Also, the programs executed by the information processing apparatuses of the first to fourth embodiments may be provided or distributed via a network such as the Internet.

[0143] The programs executed by the information processing devices of the first to fourth embodiments can cause a computer to function as each unit of the information processing device described above. In this computer, the CPU 51 can read the programs from a computer-readable storage medium onto the main storage device and execute them.

[0144] A configuration example of the embodiment will be described below. (Configuration example 1) setting a constraint condition including a first range of values ​​of a plurality of parameters included in a first model in which a plurality of explanatory variables are input and which correspond to the plurality of explanatory variables; constructing the first model by determining the plurality of parameters that satisfy the constraints; Processing section An information processing device comprising: (Configuration example 2) The processing unit setting the constraint conditions using knowledge data that is obtained in advance for at least some of the plurality of parameters and that indicates lower and upper limits of the parameter values; The information processing device according to configuration example 1. (Configuration example 3) The processing unit setting the constraint conditions using second ranges of values ​​of the plurality of parameters included in one or more second models already constructed to which the plurality of explanatory variables are input; The information processing device according to configuration example 1 or 2. (Configuration example 4) The processing unit The first range is set to a range having an upper limit of a value corresponding to a first percentile among the values ​​in the second range and a lower limit of a value corresponding to a second percentile that is lower than the first percentile. The information processing device according to configuration example 3. (Configuration Example 5) The processing unit using knowledge data indicating first lower limits and first upper limits of values ​​of the plurality of parameters, which are obtained in advance for at least some of the parameters; an upper limit being a statistical value of values ​​corresponding to the first percentile among values ​​within the first upper limit and the second range; setting, as the first range, a range having a lower limit of a statistical value of values ​​corresponding to the second percentile among values ​​within the first lower limit and the second range; The information processing device according to configuration example 4. (Configuration Example 6) The processing unit setting a range of values ​​of the parameters included in the constraint condition according to a distribution of signs of the values ​​of the plurality of parameters included in the second model; The information processing device according to configuration example 3. (Configuration Example 7) The processing unit setting the constraint conditions using ranges of values ​​of a plurality of the parameters included in one or more third models for estimation using as input at least some of the plurality of explanatory variables; The information processing device according to configuration example 1. (Configuration Example 8) The processing unit outputting output information including at least one of knowledge data indicating lower and upper limits of values ​​of at least some of the plurality of parameters, which are obtained in advance for the plurality of parameters, and lower and upper limits determined from second ranges of values ​​of the plurality of parameters included in one or more second models already constructed to which the plurality of explanatory variables are input; further setting the constraint conditions specified based on the output information; The information processing device according to configuration example 1. (Configuration Example 9) The processing unit constructing the first model by calculating the plurality of parameters using a loss function in which the loss decreases as the constraint condition is satisfied; The information processing device according to configuration example 1. (Configuration Example 10) The processing unit determining an unconstrained model, which is the first model, constructed by determining the plurality of parameters without using the constraint conditions; outputting the plurality of parameters included in the first model and the plurality of parameters included in the unconstrained model together; The information processing device according to any one of configuration examples 1 to 9. (Configuration Example 11) The processing unit outputting the plurality of parameters included in the first model together with the plurality of parameters included in one or more second models already constructed to which the plurality of explanatory variables are input; The information processing device according to any one of configuration examples 1 to 10. (Configuration Example 12) An information processing method executed by an information processing device, a step of setting a constraint condition including a first range of values ​​of a plurality of parameters corresponding to a plurality of explanatory variables included in a first model to be estimated by inputting the plurality of explanatory variables; constructing the first model by determining the plurality of parameters that satisfy the constraints; An information processing method including: (Configuration Example 13) On the computer, a step of setting a constraint condition including a first range of values ​​of a plurality of parameters corresponding to a plurality of explanatory variables included in a first model to be estimated by inputting the plurality of explanatory variables; constructing the first model by determining the plurality of parameters that satisfy the constraints; A program to execute.

[0145] 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]

[0146] 100, 100-2, 100-3, 100-4 Information processing device 101 Communication control unit 102 Acquisition Department 103, 103-2, 103-3, 103-4 Setting section 104, 104-2 Construction Department 105 Accuracy calculation section 106, 106-2, 106-3 Output control section 107-2, 107-4 Range calculation section 121 Target data storage unit 122-2, 122-4 Model data storage section 123-2 Knowledge Data Storage Unit 131 Input Devices 132 Display 200 Management System 201 Communication control unit 221 Storage section 300 Network

Claims

1. setting a constraint condition including a first range of values ​​of a plurality of parameters corresponding to a plurality of explanatory variables included in a first model for which estimation is performed as input; constructing the first model by determining the plurality of parameters that satisfy the constraints; Processing section An information processing device comprising:

2. The processing unit setting the constraint conditions using knowledge data that is obtained in advance for at least some of the plurality of parameters and that indicates lower and upper limits of the parameter values; The information processing device according to claim 1 .

3. The processing unit setting the constraint conditions using second ranges of values ​​of the plurality of parameters included in one or more second models already constructed to which the plurality of explanatory variables are input; The information processing device according to claim 1 .

4. The processing unit a range having an upper limit of a value corresponding to a first percentile among the values ​​in the second range and a lower limit of a value corresponding to a second percentile that is smaller than the first percentile, is set as the first range; The information processing device according to claim 3 .

5. The processing unit using knowledge data indicating first lower limits and first upper limits of values ​​of the plurality of parameters, which are obtained in advance for at least some of the plurality of parameters; an upper limit being a statistical value of values ​​corresponding to the first percentile among values ​​within the first upper limit and the second range; a range having a lower limit of a statistical value of values ​​corresponding to the second percentile among values ​​within the first lower limit and the second range as the first range; The information processing device according to claim 4 .

6. The processing unit setting a range of values ​​of the parameters included in the constraint condition according to a distribution of signs of the values ​​of the plurality of parameters included in the second model; The information processing device according to claim 3 .

7. The processing unit setting the constraint conditions using ranges of values ​​of a plurality of the parameters included in one or more third models for which estimation is performed by inputting at least some of the plurality of explanatory variables; The information processing device according to claim 1 .

8. The processing unit outputting output information including at least one of knowledge data indicating lower and upper limits of values ​​of at least some of the plurality of parameters, which are obtained in advance for the plurality of parameters, and lower and upper limits determined from second ranges of values ​​of the plurality of parameters included in one or more second models already constructed to which the plurality of explanatory variables are input; further setting the constraint conditions specified based on the output information; The information processing device according to claim 1 .

9. The processing unit constructing the first model by calculating the plurality of parameters using a loss function in which the loss decreases as the constraint condition is satisfied; The information processing device according to claim 1 .

10. The processing unit determining an unconstrained model, which is the first model, constructed by determining the plurality of parameters without using the constraint conditions; outputting the plurality of parameters included in the first model and the plurality of parameters included in the unconstrained model together; The information processing device according to claim 1 .

11. The processing unit outputting together the plurality of parameters included in the first model and the plurality of parameters included in one or more second models already constructed to which the plurality of explanatory variables are input; The information processing device according to claim 1 .

12. An information processing method executed by an information processing device, a step of setting a constraint condition including a first range of values ​​of a plurality of parameters corresponding to a plurality of explanatory variables included in a first model for estimation using the plurality of explanatory variables as input; constructing the first model by determining a plurality of the parameters that satisfy the constraints; An information processing method including:

13. On the computer, A first model is input with a plurality of explanatory variables and an estimation is performed. A first model is input with a plurality of explanatory variables and a constraint condition including a first range of values ​​of the parameters is set for each of the plurality of parameters corresponding to the plurality of explanatory variables. determining the constructing the first model by determining a plurality of the parameters that satisfy the constraints; A program to execute.

Images