Information Processing Apparatus, Information Processing Method, and Program

Abstract

To allow for more easily verifying validity of a model.SOLUTION: An information processing apparatus includes a storage control unit, a selection unit, and an update unit. The storage control unit stores one or more pieces of history information including identification information of models and update histories of the models into a storage unit, wherein each of the models receives input data including a plurality of variables of which degrees of influence on output data are calculated and outputs the output data and is updated by using one or more pieces of first input data. The selection unit selects a target model to be updated by using second input data, from the models identified by the identification information included in the one or more pieces of history information. The update unit updates the target model by transfer learning which uses the second input data to estimate an updated parameter with the target model as an initial value.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] Embodiments of the present invention relate to an information processing apparatus, an information processing method, and a program.

Background Art

[0002] In a machine learning model that is continuously updated, such as a prediction model and an anomaly detection model in a monitoring system of a factory or a plant, stable model updating may be required from the viewpoints of model validity verification and factor analysis. Techniques have been proposed that enable stable model updating by considering the model before updating during the learning of the machine learning model.

[0003] In the data obtained from a real monitoring system, the distribution of the data may change significantly temporarily due to, for example, a change in the operating status of manufacturing equipment or a sensor failure.

Prior Art Documents

Non-Patent Documents

[0004]

Non-Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0005] However, in the prior art, since a special period in which the data distribution changes significantly temporarily is not considered, the factors indicated by the model change greatly before and after this period, and there is a problem that it becomes difficult to verify the validity of the model or analyze the factors.

Means for Solving the Problems

[0006] The information processing apparatus according to the embodiment includes a storage control unit, a selection unit, and an update unit. The storage control unit is a model that inputs input data including a plurality of variables for which the influence degrees on the output data are respectively calculated and outputs the output data, and stores in the storage unit one or more pieces of history information including identification information of the model updated using each of one or more pieces of first input data and the history of the update of the model. The selection unit selects a target model to be updated using the second input data from the models identified by the identification information included in the one or more pieces of history information. The update unit updates the target model by transfer learning that estimates the parameters after update using the second input data with the target model as the initial value.

Brief Description of the Drawings

[0007]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Embodiments for Carrying Out the Invention

[0008] Hereinafter, with reference to the accompanying drawings, preferred embodiments of the information processing apparatus according to the present invention will be described in detail.

[0009] The information processing apparatus according to this embodiment has, for example, the following functions. This makes it possible to more easily realize the validity verification and factor analysis of the model even when the distribution of data temporarily changes significantly unintentionally. · A function of storing the models updated in the past and the history of updates (learning history) · A function of calculating the evaluation value of each of the stored models using new data · A function of selecting the optimal model from the stored models and setting it as the update target · A function of determining the period during which unintentional data is obtained temporarily

[0010] FIG. 1 is a block diagram showing an example of the configuration of an information processing system including the information processing apparatus according to this embodiment. As shown in FIG. 1, the information processing system has a configuration in which an information processing apparatus 100 and a management system 200 are connected via a network 300.

[0011] Each of the information processing apparatus 100 and the management system 200 can be configured as, for example, a server device. The information processing apparatus 100 and the management system 200 may be realized as a plurality of physically independent devices (systems), or the respective functions may be configured within one physical device. In the latter case, the network 300 may not be provided. At least one of the information processing apparatus 100 and the management system 200 may be constructed in a cloud environment.

[0012] The network 300 is, for example, a network such as a LAN (Local Area Network) and the Internet. The network 300 may be either a wired network or a wireless network. The information processing apparatus 100 and the management system 200 may transmit and receive data using a direct wired connection or wireless connection between components without going through the network 300.

[0013] The management system 200 is a system that manages the models processed by the information processing device 100 and the data used for learning (estimation) and analysis of the models. The management system 200 includes a storage unit 221 and a communication control unit 201.

[0014] The storage unit 221 stores various information used in various processes executed by the management system 200. For example, the storage unit 221 stores input data used for model estimation. The storage unit 221 can be composed of any commonly used storage medium such as a flash memory, a memory card, a RAM (Random Access Memory), an HDD (Hard Disk Drive), and an optical disk.

[0015] A model is a model that inputs input data including a plurality of variables (explanatory variables) and outputs output data (target variable) that is an inference result, and is a machine learning model that is learned (updated) by machine learning using learning input data. Each of the plurality of variables is a variable for which the degree of influence on the output data can be calculated. The model is, for example, a linear regression model, a polynomial regression model, a logistic regression model, a Poisson regression model, a generalized linear model, and a generalized additive model. The model is not limited to these.

[0016] The model is estimated by learning using input data including a target variable and explanatory variables. The target variable is, for example, a quality characteristic, a defect rate, and information indicating either a good product or a defective product. The explanatory variables are other sensor values, set values such as processing conditions, and control values.

[0017] The communication control unit 201 controls communication with external devices such as the information processing device 100. For example, the communication control unit 201 transmits input data to the information processing device 100.

[0018] Each of the above units (communication control unit 201) is implemented by, for example, one or more processors. For example, each of the above units may be implemented by causing a processor such as a CPU (Central Processing Unit) to execute a program, that is, by software. Each of the above units may be implemented by a processor such as a dedicated IC (Integrated Circuit), that is, by hardware. Each of the above units may be implemented by using a combination of software and hardware. When using a plurality of processors, each processor may implement one of the units or two or more of the units.

[0019] The information processing apparatus 100 includes a storage unit 121, an input device 122, a display 123, a communication control unit 101, a storage control unit 102, a reception unit 103, a prediction unit 104, an evaluation unit 105, a selection unit 106, an update unit 107, a generation unit 111, and a display control unit 112.

[0020] The storage unit 121 stores various information used in various processes executed by the information processing apparatus 100. For example, the storage unit 121 stores the parameters of the model updated by the update unit 107 and the learning history of the updated model. The storage unit 121 can be configured by any generally used storage medium such as a flash memory, a memory card, a RAM, an HDD, and an optical disk.

[0021] The input device 122 is a device for inputting information by a user or the like. The input device 122 is, for example, a keyboard and a mouse. The display 123 is an example of an output device that outputs information and is, for example, a liquid crystal display. The input device 122 and the display 123 may be integrated, for example, like a touch panel.

[0022] The communication control unit 101 controls communication with an external device such as the management system 200. For example, the communication control unit 101 receives input data and the like from the management system 200.

[0023] Figure 2 is a diagram showing an example of input data. The input data includes a data period, a date and time, explanatory variables, and an objective variable. The data period indicates the period (range of dates and times) during which a plurality of data (explanatory variables, objective variable) are acquired. The date and time indicates the date and time at which each of the plurality of data is acquired. As shown in Figure 2, the input data may include a plurality of explanatory variables.

[0024] Returning to Figure 1, the memory control unit 102 stores the parameters of the updated model in the storage unit 121. Figure 3 is a diagram showing an example of the parameters of the model. The model in Figure 3 is an example of a regression model having coefficients β multiplied by each of a plurality of explanatory variables as parameters.

[0025] Returning to Figure 1, the memory control unit 102 further stores in the storage unit 121 one or more pieces of history information including the identification information of the model updated using each of one or more input data (first input data) and the learning history of the model.

[0026] Each piece of history information is represented, for example, as a pair (M, H) of a model M and the learning history of the model M. Note that "M" is an example of the identification information of the model. Hereinafter, the model identified by the identification information M may be referred to as model M.

[0027] The learning history is information indicating which of the models that have been estimated or updated in the past the model M was updated with respect to. The learning history is represented, for example, by the history of the data periods of the input data used for the update. The method of representing the learning history is not limited to this. For example, the learning history may be represented by the history of the identification information of the model (target model) that was the target of the update. Further, the learning history may include both the history of the data period and the history of the identification information of the target model.

[0028] The memory control unit 102 is, for example, a set S = {(M1, H1), ···, (M N , H N)} is stored in the storage unit 121. The storage control unit 102 reads and writes the history information as necessary when selecting the target model to be updated next and when updating (learning) the model using the selected target model.

[0029] The reception unit 103 receives the input of various information. For example, the reception unit 103 receives a plurality of input data received from the management system 200 via the communication control unit 201 and the communication control unit 101. The input data includes, for example, data D = (X, Y) consisting of a pair of an explanatory variable X and an objective variable Y, and a data period h indicating the period during which the data D was acquired. When a plurality of explanatory variables are used, the explanatory variable X can be interpreted as representing, for example, a vector having each of the plurality of explanatory variables as an element.

[0030] The reception unit 103 inputs the input data D and the data period h to the prediction unit 104 and the update unit 107. The data D input to the prediction unit 104 is used for predicting the objective variable for each model included in the history information. The update unit 107 updates (learns) the parameters of the target model using, for example, the data D and the data period h.

[0031] The prediction unit 104 predicts the objective variable using the input data D (second input data) for each of one or more models identified by the identification information included in the history information. For example, the prediction unit 104 predicts the predicted value Ŷ of the objective variable Y with respect to the explanatory variable X for each model M1, ···, M N in the history information of the storage unit 121.

[0032] The evaluation unit 105 obtains an evaluation value representing the prediction accuracy of each model using the predicted value Ŷ predicted by the prediction unit 104. The evaluation value is used for the selection unit 106 to select the target model to be updated.

[0033] For example, the evaluation unit 105 calculates, for each model (model M1, ···, M NFor (), the mean squared error is calculated as an evaluation value from the target variable Y and the predicted value Ŷ obtained by the prediction unit 104. The evaluation value is not limited to the mean squared error, and may be a value calculated by other criteria such as the coefficient of determination and the mean absolute error. The evaluation value of each calculated model is input to the selection unit 106.

[0034] The selection unit 106 selects a target model to be updated from the models included in the history information. For example, the selection unit 106 selects, as the update target, a model having an evaluation value indicating higher prediction accuracy than other models.

[0035] For example, when the evaluation value is the mean squared error or the mean absolute error, the selection unit 106 selects, as the target model, the model having the smallest evaluation value. When the evaluation value is the coefficient of determination, the selection unit 106 selects, as the target model, the model having the largest evaluation value. Hereinafter, the selected target model is denoted as M best and the learning history of the target model M best is denoted as H best .

[0036] The update unit 107 updates the model. In the second and subsequent learning, the update unit 107 performs model update using transfer learning using the models learned in the past. At the first learning, since there is no model learned in the past, the update unit 107 learns the model by a method that does not use the model learned in the past.

[0037] For example, the update unit 107 updates the parameters of the target model by transfer learning that estimates the parameters of the model using the input data D with the target model selected by the selection unit 106 as the initial value. More specifically, the update unit 107 performs transfer learning using the model M best input from the selection unit 106 and the data D input from the reception unit 103, and updates the model. Let the updated model be M new . The update unit 107 adds the data period h input from the reception unit 103 to the learning history H best and updates it to H newLet it be so. The update unit 107 uses the memory control unit 102 to store the updated model and the history information (M new , H new ) in the storage unit 121.

[0038] The update unit 107 may preset learning parameters (hyperparameters) used in model learning (updating) and a threshold value (maximum number of models) indicating the maximum number of models to be stored in the storage unit 121. The maximum number of models is used, for example, in the management of the storage area of the storage unit 121 by the memory control unit 102.

[0039] The memory control unit 102 may have a function of deleting a part of the history information stored in the storage unit 121 according to a predetermined condition. For example, the memory control unit 102 performs deletion processing after updating the model so that the total number of models stored in the storage unit 121 does not become excessive. In the deletion process, the memory control unit 102 inputs the set S = {(M1, H1), ···, (M N , H N )} of the history information stored in the storage unit 121, and when the size of the set (the number of history information included in the set) is larger than the maximum number of models (an example of the condition), the oldest history information (M1, H1) is deleted. The memory control unit 102 stores the set S -1 ={(M2, H2), ···, (M N , H N )} in the storage unit 121.

[0040] As described above, the prediction unit 104 predicts the target variable for each model stored in the storage unit 121. Therefore, as the maximum number of models increases, the processing load for prediction increases. On the other hand, if the history information corresponding to a period before a period when the data distribution may temporarily change significantly is not stored, a situation may occur where an appropriate model cannot be selected. Therefore, the maximum number of models may be determined in consideration of the processing load and the length of a period during which the data distribution may temporarily change significantly.

[0041] The generation unit 111 generates visualization information for display on the display 123 or the like. For example, the generation unit 111 generates, as visualization information, attribute information representing the attributes of a model (designated model) identified by the identification information included in the history information stored in the storage unit 121 and designated by a user or the like.

[0042] For example, the reception unit 103 receives a designated model designated by a user using the input device 122 or the like. Hereinafter, the designated model is denoted as M s and the learning history of the model M s is denoted as H s .

[0043] The attribute information may be any information, but is, for example, the following information (A1) to (A4). (A1) Influence degree of each explanatory variable on the target variable (A2) Among the parameters of the designated model, the parameters that have changed with respect to the target model selected when the designated model is updated (A3) Period (history of data periods) in which each of one or more input data used for updating the designated model was obtained (A4) Out-of-scope period representing the period in which the input data was not used for updating the designated model

[0044] For example, the generation unit 111 refers to the parameters of the designated model M s and extracts the explanatory variables that contribute to the prediction of the designated model M s and generates a list of the extracted explanatory variables as attribute information (A1).

[0045] Also, the generation unit 111 refers to the learning history H s to identify the model immediately before the model M s (the model from which M s was updated). The generation unit 111 compares the parameters of the identified model with the parameters of the designated model M s to obtain the changed parameters. The generation unit 111 generates attribute information indicating the changed parameters (A2).

[0046] Also, the generation unit 111 refers to the learning history H s and generates attribute information indicating the period during which the input data used for updating the specified model was obtained (A3).

[0047] Also, the generation unit 111 refers to the learning history H s and identifies a blank period during which the input data has not been used for updating the specified model, and generates an out-of-scope period representing the identified period as attribute information (A4).

[0048] Note that during normal times when the data distribution has not changed significantly unintentionally, usually the latest model (the model learned with the input data from the most recent period) is selected as the target model. On the other hand, when the data distribution has changed significantly unintentionally, the latest model may not be selected. In such a case, one or more of the most recent periods will be blank periods during which the corresponding input data is not used for model update. Also, the learning history after model update will be a history that does not include one or more of the most recent periods. In other words, the periods included in the learning history will be discontinuous. The generation unit 111 can identify such blank periods as out-of-scope periods.

[0049] The display control unit 112 controls the display (visualization) of various information on the display 123. For example, the display control unit 112 displays the attribute information (visualization information) generated by the generation unit 111 on the display 123.

[0050] Each of the above components (communication control unit 101, memory control unit 102, reception unit 103, prediction unit 104, evaluation unit 105, selection unit 106, update unit 107, generation unit 111, and display control unit 112) is realized by, for example, one or more processors. For example, each of the above components may be realized by causing a processor such as a CPU to execute a program, that is, by software. Each of the above components may be realized by a processor such as a dedicated IC, that is, by hardware. Each of the above components may be realized by using a combination of software and hardware. When using a plurality of processors, each processor may realize one of the components or two or more of the components.

[0051] Hereinafter, an example using an information processing system for performing quality control on a manufacturing apparatus for a certain product PA will be mainly described. Product PA is a product that becomes a defective product when, for example, its concentration is less than a certain threshold. A concentration sensor value detected by a certain concentration sensor provided in the manufacturing apparatus is used for monitoring the quality of product PA.

[0052] In addition, the manufacturing apparatus includes various sensors such as a current sensor, a temperature sensor, and other concentration sensors in addition to this concentration sensor. In the present embodiment, a model is constructed that predicts a concentration sensor value (target variable) to be monitored as output data using sensor values from these sensors as input data (explanatory variables). This model is a model capable of presenting the influence degree of each input data on the prediction. By analyzing factors related to quality using the influence degree, it becomes possible, for example, to work on improving the yield. Hereinafter, an example using the Transfer Lasso (Least Absolute Shrinkage and Selection Operator) technique, which is a technique described in Non-Patent Document 1, as a model learning method will be shown.

[0053] FIG. 4 is a flowchart showing an example of the model estimation process of the embodiment. The model estimation process is a process for estimating the first model that is the basis for updating.

[0054] The update unit 107 sets the learning parameters used by the update unit 107 and the maximum number of models to be stored in the storage unit 121 (step S101). For example, in the Transfer Lasso technique, the regularization parameter and the transition parameter are set as the learning parameters.

[0055] The reception unit 103 receives the input of the initial data and the data period from the management system 200 (step S102). The initial data is data D1=(X1,Y1) that includes the concentration sensor value that becomes the target variable Y1 and other sensor values that become the explanatory variable X1 obtained within the data period h1 (for example, one month). The data format of the initial data is the same as the data format of the input data shown in FIG. 2, for example.

[0056] The update unit 107 learns a model using the input data D1 according to the set learning parameters (step S103). In the Transfer Lasso technique, the update unit 107 uses y as the target value and X as the input data of the model, and learns the coefficient β={β1,···,β p} such that y = Xβ. p is the number of elements of the explanatory variable X and the coefficient β. Each coefficient β1,···,β p corresponds to the influence degree of the corresponding explanatory variable (sensor value of each sensor) on the concentration sensor that is the target variable.

[0057] In the Transfer Lasso technique, the first model is learned by the learning method using Lasso regression. The learned model is set as the new model M1.

[0058] The update unit 107 stores the history information including the model M1 and the learning history H1 in the storage unit 121 with the learning history H1 of the model M1 being H1=[h1] (step S104). Further, the update unit 107 stores the coefficient β={β1,···,β p} and the sensor name corresponding to each coefficient in the storage unit 121 as the information (parameters) of the model M1. FIG. 3 above is an example of the parameters stored in this way.

[0059] FIG. 5 is a flowchart showing an example of the model update process of the embodiment. The model update process is a process of updating the model based on the first model estimated by FIG. 4. The model update process can be repeatedly executed for the updated model using newly obtained input data.

[0060] The reception unit 103 receives the input data D t and the data period h t from the management system 200 (step S201). The input data D t is data including the concentration sensor value that becomes the target variable Y t (for example, within one month), and the explanatory variable X t which is other sensor values. t

[0061] Next, the prediction unit 104 reads all the models M1, ···, M N and the learning histories H1, ···, H N stored in the storage unit 121 from the storage unit 121. For each of the read models, the prediction unit 104 calculates the predicted value Ŷ t of the target variable Y t which is the output data when the explanatory variable X t is input (step S202). In the Transfer Lasso technique, the predicted value Ŷ k of the model M t k (1 ≦ k ≦ N) is calculated by Ŷ t k = Xβ k .

[0062] Next, the evaluation unit 105 calculates the evaluation value of each model using the predicted value of each model (step S203). For example, when using the mean squared error as the evaluation value, the evaluation unit 105 calculates the evaluation value E k of the model M k by the following formula (1).

Equation

[0063] The selection unit 106 refers to the models M1, ···, M N and their respective evaluation values E1, ···, E N to select the model corresponding to the best evaluation value as the model M to be updated (step S204). best

[0064] The update unit 107 learns the selected target model using the input data (step S205). For example, the update unit 107 inputs the target model M best and the learning history H best corresponding to the target model M best from the selection unit 106. Also, the update unit 107 inputs the data D t =(X t, Y t ) and the data period h t from the reception unit 103. The update unit 107 performs model update based on the Transfer Lasso technique using the data D t =(X t, Y t ) and the model M best to obtain the updated model M new . Also, the update unit 107 updates the learning history to H new =[H best , h t .

[0065] The memory control unit 102 stores the history information including the updated model M new and the learning history H new in the storage unit 121 (step S206).

[0066] Next, the memory control unit 102 reads out the set of history information stored in the storage unit 121 from the storage unit 121. The memory control unit 102 determines whether the number of models included in the set of history information input from the storage unit 121 is greater than the maximum number of models (step S207). The maximum number of models is set, for example, in step S101 of FIG. 4.

[0067] ​When the number of models is greater than the maximum number of models (step S207: Yes), the memory control unit 102 deletes the oldest model and the learning history corresponding to the oldest model from the set of history information, and inputs and replaces the set of history information after deletion into the storage unit 121 (step S208).

[0068] Next, the visualization process of generating and visualizing visualization information (attribute information) will be described. FIG. 6 is a flowchart showing an example of the visualization process.

[0069] For example, the display control unit 112 displays a selection screen for selecting a model to be visualized among the models stored in the storage unit 121 on the display 123. The user selects a model to be visualized using the input device 122. Hereinafter, the selected model is designated as model M s and the learning history of the designated model M s is denoted as H s

[0070] The reception unit 103 receives the designated model M s selected (designated) as described above (step S301). Thereafter, the generation unit 111 generates attribute information (visualization information) of the designated model M s and the display control unit 112 visualizes the attribute information on the display 123 or the like.

[0071] The attribute information is, for example, the information shown in the above (A1) to (A4). Among the plurality of pieces of attribute information, the user or the like may be able to select the attribute information to be visualized. When visualizing the attribute information of (A1) to (A4), the following steps S302 to step S305 are executed respectively. The execution order of these steps is not limited to the order shown in FIG. 6. Also, for example, when not selected as the attribute information to be visualized, some of these steps may be omitted.

[0072] The generation unit 111 generates visualization information representing the degree of influence (step S302). For example, the generation unit 111 designates the model M s ​Extract the explanatory variables that contribute to the prediction. In the Transfer Lasso technique, the variables that contribute to the prediction are the variables for which the coefficient β is not zero, and the magnitude (absolute value) of the coefficient β represents the degree of influence.

[0073] FIG. 7 is a diagram showing an example of calculating the degree of influence. FIG. 7 shows an example of calculating the degree of influence when the parameters of the specified model M s are the coefficients β shown in FIG. 3. As shown in FIG. 7, for the coefficient β with a value of 0, the degree of influence may not be calculated.

[0074] Returning to FIG. 6, the generation unit 111 generates visualization information representing the change of the model (step S303). For example, the generation unit 111 s refers to the learning history H s of the specified model M s to identify the model M s-1 before the update of the specified model M. The generation unit 111 calculates the change of the specified model M s-1 with respect to the model M s . In the model based on the Transfer Lasso technique, the difference between the coefficients between the specified model M s and the model M s-1 is used as the change of the model.

[0075] The generation unit 111 refers to the learning history H s to generate visualization information indicating the period during which the input data used for the update of the specified model M s was obtained (step S304).

[0076] The generation unit 111 generates visualization information representing the out-of-scope period (step S305). For example, the generation unit 111 refers to the learning history H s to determine the non-consecutive periods, and sets the determined periods as the out-of-scope periods. FIG. 8 is a diagram showing an example of estimating the out-of-scope period. In FIG. 8, the data period where the symbol "〇" is set indicates the period during which the input data was obtained. In this example, the generation unit 111 estimates April 2020 and May 2020 as the out-of-scope periods.

[0077] The display control unit 112 visualizes the generated visualization information on the display 123 or the like (step S306). FIG. 9 is a diagram showing an example of a display screen 901 for displaying the visualization information.

[0078] Graph 911 represents the influence degree of each explanatory variable. Graph 912 represents the change of the model in the current data period (October) with respect to the previous data period (July). The change of the model is represented by, for example, the amount of change of the coefficient β for each sensor corresponding to the changed coefficient β. Graph 913 represents the change of the target variable in each period together with the learning history (history of data periods) and the out-of-scope period. Graph 914 represents the change of the target variable in the current data period.

[0079] Note that the display screen 901 in FIG. 9 is an example, and the visualization method of the visualization information is not limited to this. For example, among the graphs shown in FIG. 9, only the graph corresponding to the attribute information specified by the user or the like may be visualized.

[0080] As described above, according to the present embodiment, even when the distribution of data temporarily changes significantly unintentionally, it is possible to more easily realize the validity verification of the model, factor analysis, and the like.

[0081] Next, the hardware configuration of the information processing apparatus according to the embodiment will be described with reference to FIG. 10. FIG. 10 is an explanatory diagram showing an example of the hardware configuration of the information processing apparatus according to the embodiment.

[0082] The information processing apparatus according to the embodiment includes a control device such as a CPU 51, a storage device such as a ROM (Read Only Memory) 52 and a RAM 53, a communication I / F 54 that connects to a network and performs communication, and a bus 61 that connects each part.

[0083] The program executed by the information processing apparatus according to the embodiment is provided by being pre-embedded in the ROM 52 or the like.

[0084] The program executed by the information processing apparatus according to the embodiment may be recorded 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), a DVD (Digital Versatile Disk), etc. in an installable or executable file format and provided as a computer program product.

[0085] Furthermore, the program executed by the information processing apparatus according to the embodiment may be stored on a computer connected to a network such as the Internet and provided by being downloaded via the network. Also, the program executed by the information processing apparatus according to the embodiment may be configured to be provided or distributed via a network such as the Internet.

[0086] The program executed by the information processing apparatus according to the embodiment can cause a computer to function as each part of the above-described information processing apparatus. This computer can read a program from a computer-readable storage medium and execute it on the main storage device by the CPU 51.

[0087] Although some embodiments of the present invention have been described, these embodiments are presented by way of example and are not intended to limit the scope of the invention. These novel embodiments can be implemented in various other forms, and various omissions, replacements, and changes can be made without departing from the gist of the invention. These embodiments and their modifications are included in the scope and gist of the invention, and are included in the invention described in the claims and its equivalent scope.

Explanation of Reference Numerals

[0088] 100 Information processing apparatus 101 Communication control unit 102 Storage control unit 103 Reception unit 104 Prediction unit 105 Evaluation Unit 106 Selection Unit 107 Update Unit 111 Generation Unit 112 Display Control Unit 121 Memory Unit 122 Input Device 123 Display 200 Management System 201 Communication Control Unit 221 Memory Unit 300 Network

Claims

1. A model that inputs input data including a plurality of variables for which the degree of influence on output data is calculated respectively, and outputs the output data, the storage control unit storing in a storage unit one or more pieces of history information including identification information of the model updated using each of one or more pieces of first input data input in mutually different data periods, a prediction unit that predicts the output data using second input data for each of one or more of the models identified by the identification information included in the history information, an evaluation unit that obtains an evaluation value representing the accuracy of prediction of each of the models based on the output data, a selection unit that selects, as a target model to be updated using the second input data, a model having the evaluation value indicating that the prediction accuracy is higher than that of other models, an update unit that updates the target model by transfer learning that estimates parameters after update using the second input data with the target model as an initial value, An information processing apparatus comprising:

2. The model is a regression model that inputs input data including a plurality of explanatory variables and outputs output data that is a target variable, The evaluation value is any one of a mean squared error, a coefficient of determination, and a mean absolute error, The information processing apparatus according to claim 1.

3. When the number of the history information is larger than a threshold value, the storage control unit deletes a part of the history information stored in the storage unit, The information processing apparatus according to claim 1 or 2.

4. a generation unit that generates attribute information representing an attribute of a designated model that is a model identified by the identification information included in the designated history information among the history information, a display control unit that visualizes the attribute information, further comprising: The information processing apparatus according to any one of claims 1 to 3.

5. The generation unit generates the degree of influence as the attribute information, The information processing apparatus according to claim 4.

6. The generation unit generates the attribute information indicating a parameter that has changed with respect to the target model selected when updating the designated model among the parameters of the designated model, The information processing apparatus according to claim 4.

7. The generation unit generates the attribute information indicating the data period, The information processing apparatus according to claim 4.

8. The history information further includes a history of the data period of the first input data used for updating the model, The generation unit generates the attribute information indicating the out-of-scope period representing the data period during which the first input data was not used for updating the designated model, based on the history of the data period included in the history information. The information processing apparatus according to claim 4.

9. An information processing method executed by an information processing apparatus, a storage control step of storing, in a storage unit, one or more pieces of history information including identification information of a model updated using each of one or more first input data input in mutually different data periods, the model being one that inputs input data including a plurality of variables for which influence degrees on output data are respectively calculated and outputs the output data; a prediction step of predicting the output data using second input data for each of one or more of the models identified by the identification information included in the history information; an evaluation step of obtaining an evaluation value representing the prediction accuracy of each of the models based on the output data; a selection step of selecting, as a target model to be updated using the second input data, a model having the evaluation value indicating that the prediction accuracy is higher than that of other models; an update step of updating the target model by transfer learning for estimating parameters after update using the second input data, with the target model as an initial value; An information processing method including the above steps.

10. A program for causing a computer to execute a storage control step of storing, in a storage unit, one or more pieces of history information including identification information of a model updated using each of one or more first input data input in mutually different data periods, the model being one that inputs input data including a plurality of variables for which influence degrees on output data are respectively calculated and outputs the output data; a prediction step of predicting the output data using second input data for each of one or more of the models identified by the identification information included in the history information; an evaluation step of obtaining an evaluation value representing the prediction accuracy of each of the models based on the output data; a selection step of selecting, as a target model to be updated using the second input data, a model having the evaluation value indicating that the prediction accuracy is higher than that of other models; an update step of updating the target model by transfer learning for estimating parameters after update using the second input data, with the target model as an initial value; and execute the above steps.

Images