Data analysis support device and data analysis support method

Abstract

To provide a data analysis support device for facilitating analysis of data about product manufacturing.SOLUTION: A data analysis support device includes a storage part for storing a data set and a prediction model, an input part and a control part. The data set includes an explanatory variable for specifying at least one parameter for manufacturing a product and an objective variable for specifying at least one parameter showing the property of the product. The prediction model associates the explanatory variable with the objective variable. The input part accepts an input of input information corresponding to the explanatory variable. The control part acquires output information showing the property of the product by using the prediction model on the basis of the input information. The control part generates distribution map data showing a relation between the explanatory variable and the objective variable of the data set, plots the input information and the output information on a distribution map shown by the distribution map data, and causes a display device to display the distribution map obtained by plotting the input information and the output information.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to a data analysis support device and a data analysis support method. [Background technology]

[0002] Patent Document 1 discloses a method for displaying two types of data: at least one parameter determined as a design variable among parameters that define a structure and a material that constitutes the structure, and at least one parameter determined as a characteristic value among parameters that define a structure and a material that constitutes the structure. When displaying Pareto solutions as a scatter diagram in an objective function space, the computer disclosed in Patent Document 1 changes at least one of the color, type, and size of a symbol representing the value of the design variable in the scatter diagram according to the value of the design variable. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Patent No. 6248402 Summary of the Invention [Problem to be solved by the invention]

[0004] The present invention provides a data analysis support device and a data analysis support method that facilitate the analysis of data relating to the manufacture of products. [Means for solving the problem]

[0005] A data analysis support device according to one aspect of the present invention comprises: A storage unit that stores a data set including explanatory variables that define at least one parameter for manufacturing a product and a response variable that defines at least one parameter that indicates a property of the product, and a prediction model that associates the explanatory variables with the response variable; an input unit that receives input information corresponding to the explanatory variables; a control unit that obtains output information indicating a property of the product using the prediction model based on the input information, The control unit is generating distribution map data showing a relationship between the explanatory variables and the objective variable of the data set; Plotting the input information and the output information on a distribution map indicated by the distribution map data; The distribution map in which the input information and the output information are plotted is displayed on a display device.

[0006] In one embodiment, the objective variable defines a plurality of parameters indicative of properties of the product; The distribution map data may include information indicating a relationship between the plurality of parameters of the response variable.

[0007] In one aspect, the control unit is generating a histogram of the response variable; The histogram may be displayed on the display device together with the distribution map.

[0008] In one aspect, the prediction model is a trained model trained to output the objective variable in response to an input of the explanatory variable, The control unit may acquire the output information output from the prediction model by inputting the input information to the prediction model.

[0009] In one aspect, the control unit is determining trend information indicative of a trend of change in the objective variable when the explanatory variable changes based on the prediction model; An image indicating the trend information may be displayed superimposed on the distribution map.

[0010] In one aspect, the control unit may cause the display device to display reference information indicating at least one of a range of the explanatory variables and a range of the objective variable of the data set.

[0011] In one aspect, the reference information includes information indicating a range of the explanatory variable, The control unit is inputting reference input information within the range of the explanatory variables of the reference information into the prediction model, thereby obtaining reference output information output from the prediction model; The reference input information and the reference output information may be displayed on the display device.

[0012] In one aspect, the control unit is an external data set including the explanatory variables and the objective variable, and an external model associating the explanatory variables with the objective variable, are externally acquired via the input unit; When the external data set or the external model is acquired via the input unit, determining whether class information of the acquired external data set or the external model matches class information of the data set or the prediction model; When it is determined that the class information of the external dataset or the external model matches the class information of the dataset or the predictive model, the external dataset or the external model may be stored in the storage unit.

[0013] A data analysis support method according to one aspect of the present invention is a data analysis support method executed by a control unit, comprising: A step of preparing a data set including explanatory variables defining at least one parameter for manufacturing a product and a response variable defining at least one parameter indicating a property of the product, and a prediction model relating the explanatory variables and the response variable; acquiring input information corresponding to the explanatory variables; using the predictive model to obtain output information indicative of a property of the product based on the input information; generating distribution map data showing a relationship between the explanatory variables and the objective variable of the data set; Plotting the input information and the output information on a distribution map represented by the distribution map data; and displaying, on a display device, the distribution map in which the input information and the output information are plotted. Effect of the Invention

[0014] According to the data analysis support device and data analysis support method of the present invention, it is possible to facilitate the analysis of data relating to the manufacture of products. [Brief description of the drawings]

[0015] [Figure 1] FIG. 1 is a block diagram showing a configuration of a data analysis support system according to an embodiment of the present invention. [Diagram 2] Table showing an example dataset [Diagram 3] 1 is a flowchart illustrating a prediction operation of a data analysis support method according to an embodiment of the present invention. [Figure 4] FIG. 13 is a schematic diagram showing an example of a model selection screen; [Diagram 5] FIG. 13 is a schematic diagram showing an example of a prediction execution screen. [Figure 6] FIG. 13 is a schematic diagram showing an example of a prediction execution screen. [Figure 7] FIG. 13 is a schematic diagram illustrating a first graph screen showing a graph between objective variables and a histogram; [Figure 8] FIG. 13 is a schematic diagram illustrating a second graph screen showing a distribution diagram; [Figure 9] FIG. 13 is a schematic diagram illustrating a third graph screen including a trend line. [Figure 10] FIG. 1 is a schematic diagram illustrating an example of a process for generating a trend line; [Figure 11] FIG. 1 is a schematic diagram illustrating an example of a process for generating a trend line; [Figure 12] FIG. 1 is a schematic diagram illustrating an example of a process for generating a trend line; [Figure 13] 1 is a flowchart illustrating a search operation of a data analysis support method according to an embodiment of the present invention. [Figure 14] FIG. 13 is a schematic diagram showing an example of a search setting screen for setting search conditions. [Figure 15]FIG. 13 is a schematic diagram showing an example of a search result screen showing a search result; [Figure 16] 11 is a flowchart illustrating an external information import process according to another embodiment of the present invention; DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0016] Hereinafter, an embodiment of the present invention will be described with reference to the drawings as appropriate. The following description is merely an example of an embodiment for embodying the technical idea of ​​the present invention, and the technical idea of ​​the present invention is not limited to the following embodiment.

[0017] 1. Configuration 1-1.Hardware configuration 1 is a block diagram showing a configuration of a data analysis support system 1 according to an embodiment of the present invention. The data analysis support system 1 supports the analysis of data related to the manufacture of products. One example of data analysis is the prediction of parameters that indicate the properties of a product.

[0018] The data analysis support system 1 includes a data analysis support device 10 and a display device 20. The data analysis support device 10 includes a control unit 11, a storage unit 12, an input interface (IF) 13, and an output interface (IF) 14.

[0019] The control unit 11 includes an arithmetic circuit that cooperates with software to realize a predetermined function, and performs information processing to realize the functions of the data analysis support device 10. Such information processing is realized, for example, by the control unit 11 operating according to instructions of a program stored in the storage unit 12. For example, the control unit 11 operates as an arithmetic unit 111 and a graph generation unit 112. The control unit 11 may include various semiconductor integrated circuits, such as a CPU (Central Processing Unit), an MPU (Micro Processing Unit), and an FPGA (Field Programmable Gate Array), instead of or as the arithmetic circuit. The data analysis support method of this embodiment may also be executed by distributed computing.

[0020] Input interface 13 is an interface circuit that connects data analysis support device 10 to an input device such as a mouse, a touch panel, or a keyboard in order to input various information to data analysis support device 10. Input interface 13 may be a communication circuit that performs data communication in accordance with an existing wired communication standard or wireless communication standard.

[0021] The output interface 14 is an interface circuit that connects the data analysis support device 10 to an output device in order to output information from the data analysis support device 10. Such an output device is, for example, an external device such as a display device 20 or a server device. The output interface 14 may be a communication circuit that performs data communication in accordance with an existing wired communication standard or wireless communication standard. The input interface 13 and the output interface 14 may be realized by similar hardware.

[0022] The input interface 13 and the output interface 14 may be configured as an integrated unit, rather than as separate interfaces as in the example of FIG.

[0023] The display device 20 includes a device capable of displaying information, such as a liquid crystal display or an organic EL display.

[0024] The storage unit 12 is a recording medium for recording various information including programs necessary for implementing the functions of the data analysis support device 10. The storage unit 12 may store a data set 121 and a prediction model 122, which will be described later. The storage unit 12 is implemented, for example, by a flash memory, a semiconductor storage device such as a solid state drive (SSD), a magnetic storage device such as a hard disk drive (HDD), or other recording media, either alone or in combination. The storage unit 12 may include a volatile memory such as a random access memory (RAM).

[0025] The storage unit 12 may be any of an internal type, an external type, and a NAS (network-attached storage) type. The storage unit 12 may be a storage implemented on a cloud network, or may be a specific recording server with limited access. A unique ID may be assigned to each piece of information stored in the storage unit 12. This allows the necessary information to be easily specified using the ID as a key during prediction. The storage unit 12 may store association information for each ID. By using the ID association information, the control unit 11 can verify the effect of the setting conditions for data prediction on the prediction result.

[0026] 1-2. Predictive models and data sets The prediction model 122 is a model that associates explanatory variables with dependent variables, and predicts the dependent variables based on the explanatory variables.

[0027] The explanatory variables define at least one parameter for manufacturing a product such as a steel plate. An example of the explanatory variables is the amount and mixing ratio of ingredients in the material of the product. An example of the explanatory variables is the manufacturing conditions in the manufacturing process such as heat treatment and rolling.

[0028] The objective variable defines at least one parameter that is indicative of a property of the product. Examples of objective variables are indices such as tensile strength, elongation at break, hole expansion ratio, electrical conductivity, and corrosion resistance of the product.

[0029] In this embodiment, the prediction model 122 is a trained model trained to output a response variable in response to an input of an explanatory variable. The control unit 11 inputs the explanatory variables to the prediction model 122 to obtain a prediction result of the response variable output from the prediction model 122.

[0030] The above-mentioned learning is performed by executing a learning process using a learning dataset 121 for the model. Such learning process is executed by the control unit 11 or another information processing device using techniques such as deep learning, Gaussian process, k-nearest neighbor method, decision tree, random forest, gradient boosting, etc. The learning dataset 121 is a combination in which at least one of various manufacturing parameters for manufacturing a product is associated with a predetermined property of a product actually manufactured using the manufacturing parameter. The predetermined property of the product is obtained, for example, by performing a predetermined test on the product.

[0031] 2 is a table showing an example of the data set 121. The data set 121 includes values ​​of each parameter of the design variables (X1, X2, X3,...) and each parameter of the objective variables (Y1, Y2,...). The design variables and objective variables are not limited to numerical values, and may include rank information such as A, B, C (see X4 in FIG. 2), information for distinguishing manufacturing equipment, manufacturing processes, etc. In this embodiment, the prediction model 122 and the data set 121 are linked and stored in the storage unit 12.

[0032] 2. Predictive behavior 2-1. Overview

[0033] In order to enable a user to efficiently design a product based on a predictive model and input data to the predictive model, a function for supporting a user's understanding, operation, etc. is required. Such a support function is required to provide interpretability that allows a user to understand the operating tendency, analysis results, etc. of a predictive model even if the user is not familiar with handling predictive models or data analysis. In addition, a support function for data analysis is required to have high reliability for the results of data analysis. Reliability is, for example, an index indicating the likelihood of an analysis result. In particular, it is required to achieve both reliability and interpretability. Therefore, in this embodiment, a method is proposed that can easily provide a user with information useful for design by achieving both reliability and interpretability for the output result obtained using the predictive model 122.

[0034] 3 is a flowchart illustrating a prediction operation of the data analysis support method according to the present embodiment. The process of FIG. 3 is executed by the control unit 11 of the data analysis support device 10, for example.

[0035] 2-2. Enter prediction conditions First, the control unit 11 acquires prediction conditions (S11). The prediction conditions include the type of prediction model 122, the number of points (prediction points) that the user wants to predict, and values ​​of input information corresponding to explanatory variables. The prediction conditions are input by the user using an input device such as a mouse, a touch panel, or a keyboard, and are input to the control unit 11 via the input interface 13.

[0036] An example of a prediction condition setting means will be described with reference to FIGS.

[0037] Setting the prediction conditions includes selecting a prediction model using a model selection screen. Fig. 4 is a schematic diagram showing an example of the model selection screen 21. The model selection screen 21 in Fig. 4 is output to the display device 20 via the output interface 14 by the control unit 11, and is displayed on the display device 20.

[0038] The storage unit 12 may store a plurality of prediction models. When a plurality of prediction models are stored in the storage unit 12, the user selects a desired prediction model on the model selection screen 21 of FIG. 4. In the example shown in FIG. 4, the user can select any one of models A to G. The user can also search for a desired prediction model by inputting a keyword in the tag search window 211. When the user selects the prediction button 212 corresponding to the desired prediction model by an operation such as clicking or tapping, the screen transitions to a prediction execution screen 322a of FIG. 5.

[0039] Fig. 5 is a schematic diagram showing an example of a prediction execution screen 322a. The model field 221 of the prediction execution screen 322a displays information about the prediction model 122 selected on the model selection screen 21 of Fig. 4. The information about the prediction model 122 displayed in the model field 221 includes, for example, the model ID, the model name, the method (technology) applied to the model, explanatory variables, and objective variables.

[0040] For the prediction point, the user can input the value of each parameter of the explanatory variable as input information in the prediction point field 222 on the prediction execution screen 322a.

[0041] The prediction execution screen 322a includes a reference information field 223. The reference information field 223 displays the range of each parameter of the explanatory variable of the learning dataset 121 used in learning the prediction model 122. The range of each parameter of the explanatory variable includes, for example, the minimum value, maximum value, average value, etc. of each parameter of the explanatory variable.

[0042] The user can determine the value to be input into the prediction point field 222 by referring to the range of each parameter of the explanatory variable displayed in the reference information field 223. It is known that in a machine learning model, the prediction accuracy tends to be worse when an input value outside the range of the learning data (extrapolation field) is input than when an input value within the range (interpolation field) is input. In this embodiment, the user can compare the input value with the range of each parameter of the explanatory variable of the learning dataset 121 to check whether the input value is within the interpolation field. In this way, the reference information field 223 can assist the user in inputting each parameter of the explanatory variable.

[0043] The control unit 11 may input reference input information within the range of each parameter of the explanatory variables displayed in the reference information field 223 to the prediction model 122, and acquire the objective variable output from the prediction model 122 accordingly as reference output information. In this case, the control unit 11 may cause the display device 20 to display the reference input information and the reference output information. This allows the user to efficiently select an input value to the prediction model 122 for obtaining a desired output value by referring to the reference input information and the reference output information.

[0044] 2-3. Prediction 3, the control unit 11 executes prediction using the prediction model 122 according to the prediction conditions acquired in step S11, and obtains a prediction result (S12). The prediction process in step S12 includes predicting a response variable using explanatory variables and the prediction model 122. The prediction process in step S12 is executed by the calculation unit 111 of the control unit 11, for example.

[0045] For example, the user inputs the values ​​of each parameter of the explanatory variables in the prediction point field 222 as shown in the prediction execution screen 322b of Fig. 6, and presses the prediction execution button 224 by an operation such as clicking or tapping. In response to this, the control unit 11 inputs the explanatory variables input in the prediction point field 222 to the prediction model 122, and obtains the output from the prediction model 122 as the objective variable. The control unit 11 causes the objective variable, which is the prediction result by the prediction model 122, to be displayed in the prediction point field 222 as shown in Fig. 6.

[0046] 2-4.Graph Generation 3, the control unit 11 generates a graph based on the prediction result obtained in step S12 and the data set 121 (S13). The process of step S13 is executed by the graph generation unit 112 of the control unit 11, for example.

[0047] 6, explanatory variables and a response variable as a prediction result are displayed in a prediction point field 222. When a user presses a graph display button 225, each prediction point in the prediction point field 222 is displayed as a graph as described below. Of the prediction points in the prediction point field 222, only prediction points with a checked check box may be displayed as a graph as shown in FIG.

[0048] The graph generated in step S13 includes, for example, a graph between objective variables showing the relationship between the parameters of the objective variables of the learning dataset 121 and the relationship between the parameters of the objective variables of the prediction results obtained in step S12. In addition, the graph generated in step S13 includes, for example, a histogram for each parameter of the objective variables in the learning dataset 121.

[0049] FIG. 7 is a schematic diagram illustrating an example of the first graph screen 23 showing such inter-objective variable graphs 231, 232 and histograms 233, 234.

[0050] The horizontal axis of the objective variable graph 231 is the first parameter Y1 of the objective variable, and the vertical axis is the second parameter Y2 of the objective variable. Object variables included in the data set 121 are indicated by black dots in the objective variable graph 231. When the user selects an objective variable in the objective variable graph 231 by an operation such as clicking or tapping, the values ​​of the parameters Y1 and Y2 of the selected objective variable and the values ​​of the parameters X1 to X4 of the explanatory variables corresponding to the selected objective variable are displayed in the variable display field 235.

[0051] In the graph between objective variables 231, objective variables 231a, 231b, and 231c of the prediction results obtained in step S12 are further plotted. The objective variables 231a, 231b, and 231c correspond to the objective variables of each prediction point displayed in the prediction point column 222 of Fig. 6. In the illustrated example, the objective variables 231a, 231b, and 231c plotted in the graph between objective variables 231 of Fig. 7 correspond to the prediction points titled Demo a, Demo b, and Demo c in the prediction point column 222 of Fig. 6, respectively.

[0052] The horizontal axis of the objective variable graph 232 is Y1, and the vertical axis is Y2. The display form of the objective variable graph 232 is similar to that of the objective variable graph 231 described above. The objective variable graph 232 further plots objective variables 232a, 232b, and 232c of the prediction results obtained in step S12.

[0053] 7, in a parameter selection field 236. When the user presses a graph drawing button 237, a graph between objective variables and a histogram related to the parameter selected in the parameter selection field 236 are displayed on the first graph screen 23.

[0054] The user can grasp the distribution of the objective variables in the data set 121 by looking at the objective variable graphs 231, 232 displayed on the first graph screen 23 in FIG. 7. For example, in material data of a product, there are cases where a plurality of objective variables are in a trade-off relationship. In such a case, the user can grasp the relationship between the objective variables from the objective variable graphs 231, 232. In addition, the user can more accurately predict the likelihood of improvement in the objective variable results by selecting various explanatory variables by checking where the prediction results obtained in step S12 are located on the objective variable graphs 231, 232.

[0055] 7 shows the distribution of the first parameter Y1 of the response variable, and the histogram 234 shows the distribution of the second parameter Y2 of the response variable.

[0056] The user can grasp the reliability of the prediction result obtained in step S12 by looking at the histograms 233, 234 displayed on the first graph screen 23 in Fig. 7 together with the graphs between objective variables 231, 232. For example, if the prediction result obtained in step S12 corresponds to a position with a high frequency in the histograms 233, 234, it can be said that the reliability of the prediction result is high. In addition, the user can visually grasp the value range of the objective variable, explanatory variables, etc., for which the prediction model 122 is good at making accurate predictions, from the bias of the parameters in the learning dataset 121.

[0057] Furthermore, the graph generated in step S13 includes a distribution diagram indicated by distribution diagram data indicating the relationship between explanatory variables and objective variables of the learning dataset 121. Fig. 8 is a schematic diagram illustrating a second graph screen 24 showing such a distribution diagram. In the example of Fig. 8, the second graph screen 24 shows distribution diagrams 241-1, 241-2, 242-1, 242-2, 243-1, 243-2, 244-1, and 244-2.

[0058] The horizontal axis of the distribution maps 241-1 and 241-2 is X1, the horizontal axis of the distribution maps 242-1 and 242-2 is X2, the horizontal axis of the distribution maps 243-1 and 243-2 is X3, and the horizontal axis of the distribution maps 244-1 and 244-2 is X4. The vertical axis of the distribution maps 241-1, 242-1, 243-1, and 244-1 is the first parameter Y1 of the response variable. The vertical axis of the distribution maps 241-2, 242-2, 243-2, and 244-2 is the second parameter Y2 of the response variable.

[0059] 8, in a parameter selection field 246. When the user presses a graph drawing button 247, a distribution diagram for the parameter selected in the parameter selection field 246 is displayed on the second graph screen 24.

[0060] In each of the distribution diagrams 241-1, 241-2, 242-1, 242-2, 243-1, 243-2, 244-1, and 244-2, points indicating the prediction results obtained in step S12 are plotted, similar to the graphs between objective variables 231 and 232 in Fig. 7. In the illustrated example, each distribution diagram shows symbols corresponding to the prediction points titled demo a, demo b, and demo c in the prediction point column 222 in Fig. 6.

[0061] The second graph screen 24 may display an estimated drawing time 248. The estimated drawing time 248 is displayed, for example, as shown in FIG. 8 using a progress bar and text, but is not limited thereto, and may be displayed using either a progress bar or text. The estimated drawing time 248 allows the user to check an estimate of the time required for drawing. The estimated drawing time 248 also allows the user to know the progress of drawing the distribution map.

[0062] 2-5. Trend lines 3, the control unit 11 generates a trend line (S14). The trend line is an example of trend information that indicates the trend of change in the objective variable when the explanatory variable changes. The trend line is generated based on the data set 121.

[0063] 9 is a schematic diagram illustrating the third graph screen 24a. In the third graph screen 24a in FIG. 9, trend lines are superimposed on each distribution diagram in the second graph screen 24 in FIG.

[0064] Trend lines 11a, 11b, and 11c of distribution chart 241-1 are trend lines corresponding to the prediction points titled demo a, demo b, and demo c in the prediction points column 222 of FIG. 6. Similarly, trend lines 12a, 12b, and 12c of distribution chart 241-2 are trend lines corresponding to the prediction points titled demo a, demo b, and demo c. Trend lines 21a, 21b, and 21c of distribution chart 242-1 are trend lines corresponding to the prediction points titled demo a, demo b, and demo c. Trend lines 22a, 22b, and 22c of distribution chart 242-2 are trend lines corresponding to the prediction points titled demo a, demo b, and demo c. Trend lines 31a, 31b, and 31c of distribution chart 243-1 are trend lines corresponding to the prediction points titled demo a, demo b, and demo c. Trend lines 32a, 32b, and 32c in plot 243-2 are trend lines corresponding to forecast points entitled demo a, demo b, and demo c, respectively.

[0065] By checking the trend lines on the third graph screen 24a, the user can understand how each objective variable changes when each explanatory variable is changed. For example, by checking the trend lines on the distribution diagram 241-1 and the distribution diagram 242-1, the user can realize that if the objective variable Y1 is to be increased in the prediction model, the explanatory variables X1 and / or X2 should be increased. For example, the user can effectively develop materials by utilizing this realization in material design.

[0066] An example of the process for generating trend lines will be described with reference to Figs. 10 to 12. As shown in Fig. 10, the control unit 11 generates a distribution diagram for the prediction points selected in the prediction point column 222 of the prediction execution screen 322b. The process for generating this distribution diagram is the same as that of step S13. The horizontal axis of the distribution diagrams 245-1 and 245-2 shown in Fig. 10 is X5.

[0067] We will now explain the process of generating trend lines showing the tendency of changes in the response variables Y1 and Y2 when the explanatory variable X1 changes, for points 241-1a and 241-2a on the X1-Y2 plane (distribution chart 241-1) corresponding to the predicted point 222a in the predicted point column 222.

[0068] First, the control unit 11 selects points 241-1a and 241-2a. As shown in Fig. 10, the parameters indicating the predicted point 222a are (X1, X2, X3, X4, X5, Y1, Y2) = (0.05, 0.05, 0.05, 0, 0.05, 1.883, 0.534). Therefore, the coordinates of the point 241-1a are (X1, Y1) = (0.05, 1.883), and the coordinates of the point 241-2a are (X1, Y2) = (0.05, 0.534).

[0069] Next, the control unit 11 obtains data sets D1 to Dn by varying the parameter X1 of the explanatory variable of the points 241-1a and 241-2a from the minimum value to the maximum value at predetermined intervals as shown in Fig. 11. In the data sets D1 to Dn, the parameters X2 to X5 other than X1 are fixed values.

[0070] In the example of FIG. 11, the minimum value is 0, the maximum value is 1, and the predetermined interval is 0.01. However, this embodiment is not limited to this, and the minimum value of X1 may be set to the minimum value of the parameter X1 of the explanatory variable in the learning dataset 121, and the maximum value of X1 may be set to the maximum value of the parameter X1 of the explanatory variable in the dataset 121. Alternatively, the minimum value of X1 may be set to a value smaller than the parameter X1 of the explanatory variable in the dataset 121 by a first predetermined value, and the maximum value of X1 may be set to a value larger than the parameter X1 of the explanatory variable in the dataset 121 by a second predetermined value. The first predetermined value and the second predetermined value may be the same or different. In addition, the above-mentioned predetermined interval may be always constant, but may not be constant.

[0071] Next, the control unit 11 inputs each of the data sets D1 to Dn into the prediction model 122, and obtains objective variables Y11 to Y1n and Y21 to Y2n, which are output results.

[0072] Next, the control unit 11 draws trend lines on the distribution charts 241-1 and 241-2 based on the objective variables Y11 to Y1n and Y21 to Y2n.

[0073] Fig. 12 is a schematic diagram illustrating a trend line drawn in this manner. In Fig. 12, the trend line is superimposed on each distribution diagram of the second graph screen 24 in Fig. 10. The control unit 11 can draw the trend line 11a based on the value of the explanatory variable X1 and the objective variables Y11 to Y1n in the data sets D1 to Dn. The control unit 11 may simply plot the objective variables Y11 to Y1n on the distribution diagram 241-1, or may draw the trend line 11a as an approximation curve based on the objective variables Y11 to Y1n.

[0074] Similarly, the control unit 11 plots a trend line 12a based on the values ​​of the explanatory variable X1 and the objective variables Y21 to Y2n in the data sets D1 to Dn.

[0075] By the same process as above, the control unit 11 can draw trend lines 31a and 32a showing the tendency of change in the objective variables Y1 and Y2 when the explanatory variable X3 changes. The control unit 11 can draw trend lines 51a and 52a showing the tendency of change in the objective variables Y1 and Y2 when the explanatory variable X5 changes.

[0076] The control unit 11 can draw trend lines 11b, 11c, 12b, 12c, 31b, 31c, 32b, 32c, 51b, 51c, 52b, and 52c illustrated in FIG. 12 by performing the same processing as described above for prediction points other than prediction point 222a in FIG.

[0077] 2-6.Display 3, the control unit 11 causes the display device 20 to display the graph generated in step S13 and the trend line generated in step S14 (S15). For example, the control unit 11 transmits image data of the graph and the trend line to the display device 20 via the output interface 14, and the display device 20 displays an image indicated by the received image data.

[0078] 3.Exploration behavior 3-1. Overview Fig. 13 is a flowchart illustrating a search operation of the data analysis support method according to the present embodiment. In the search operation, a prediction model 122 is used to search for explanatory variables that output a target variable input by a user as a search target. Alternatively, the control unit 11 may cause a search model prepared for the search to execute the search instead of the prediction model 122. The process of Fig. 13 is executed by, for example, the control unit 11 of the data analysis support device 10.

[0079] 3-2. Enter search conditions First, the control unit 11 acquires search conditions (S21). The search conditions include the type of the prediction model 122, the value of the input information (search target) corresponding to the objective variable, the number of searches, etc. The search conditions are input by a user using an input device such as a mouse, a touch panel, or a keyboard, and are input to the control unit 11 via the input interface 13.

[0080] An example of a means for setting search conditions will be described with reference to Fig. 14. Fig. 14 is a schematic diagram showing an example of a search setting screen 25 for setting search conditions. A user can input a desired value of a target variable as a search target in a search target field 251 of the search setting screen 25. For example, a user specifies a search target by inputting a lower limit value and an upper limit value, as in Y1 in Fig. 14.

[0081] Alternatively, the user may set a goal such that the parameter of a certain objective variable is maximized (max), as in Y2 in FIG. 14. To do so, the user selects "max" in the goal setting field of the search goal field 251. When "max" is selected as the goal setting, the control unit 11 searches for explanatory variables that maximize the parameter of the target objective variable (Y2 in FIG. 14). The user may also set a goal such that the parameter of a certain objective variable is minimized (min). When "min" is selected as the goal setting, the control unit 11 searches for explanatory variables that minimize the parameter of the target objective variable.

[0082] Alternatively, nothing may be set in the target setting field of the search target field 251. For example, a null value may be set in the target setting field of the search target field 251. In this case, information on the parameter of the objective variable for which a null value is set in the target setting field does not have any effect, such as limiting the search results.

[0083] The user can specify the range of each parameter of the explanatory variable in the explanatory variable field 252 of the search setting screen 25. Points within the specified range are obtained as the search result.

[0084] In the search count field 253 of the search setting screen 25, the user can set the number of times to perform a search. If the search is set to be performed two or more times, points within a predetermined distance from points that have been searched up to that point may be excluded from the search targets in the second and subsequent searches. This makes it possible to prevent similar search results from being repeatedly output.

[0085] The prediction model to be used in the search is selected using a model selection screen similar to model selection screen 21 shown in FIG.

[0086] 3-3.Exploration A search is executed when the user presses the search execution 254 button (see FIG. 14) on the search setting screen 25. The control unit 11 executes a search using the prediction model 122 according to the search conditions acquired in step S21, and obtains a search result (S22 in FIG. 13).

[0087] 15 is a schematic diagram showing an example of a search result screen 26 showing the search results obtained in step S22. Information about the prediction model 122 selected by the user and used in the search is displayed in a model field 261 of the search result screen 26. The information about the prediction model 122 displayed in the model field 261 includes, for example, the model name, the search target, the explanatory variables, the acquisition function, the optimization method, the number of searches, the update date and time, and the like.

[0088] The number of search results corresponding to the number of searches is displayed in the search result field 262 of the search result screen 26. In Fig. 15, the number of searches is set to three, so three search results are displayed in the search result field 262.

[0089] The setting value display field 263 of the search result screen 26 displays the condition of the objective variable as the search target set in the search target field 251 of the search setting screen 25 of Figure 14, and the range of each parameter of the explanatory variable set in the explanatory variable field 252 of the search setting screen 25 of Figure 14.

[0090] 3-4.Graph Generation When the user presses the graph display button 264, the control unit 11 generates a graph (S23 in FIG. 13). The graph generated in step S23 is similar to the objective variable graphs 231 and 232 and the histograms 233 and 234 shown in FIG.

[0091] If the user selects “Range” for a particular objective variable in the goal setting field of the search target field 251 in FIG. 14, the control unit 11 may superimpose a reference line representing the target range of the objective variable on the graph showing the objective variable generated in step S23.

[0092] The graph generated in step S23 may include a distribution diagram such as that shown in FIG.

[0093] 4. Summary As described above, the data analysis support device 10 according to the present embodiment includes the storage unit 12 that stores the data set 121 and the prediction model 122, the input interface 13 that is an example of an input unit, and the control unit 11. The data set 121 includes explanatory variables that define at least one parameter for manufacturing a product, and a response variable that defines at least one parameter that indicates the properties of the product. The prediction model 122 associates the explanatory variables with the response variables. The input interface 13 accepts input of input information corresponding to the explanatory variables (S11). The control unit 11 obtains output information that indicates the properties of the product using the prediction model 122 based on the input information (S12). The control unit 11 generates distribution map data that indicates the relationship between the explanatory variables and the response variables of the data set 121, and plots the input information and the output information on the distribution map indicated by the distribution map data (S13). The control unit 11 causes the display device 20 to display the distribution map on which the input information and the output information are plotted (S15).

[0094] The data analysis support device 10 supports the analysis of data related to the manufacture of products. Specifically, for example, the data analysis support device 10 displays a distribution diagram in which input information and output information are plotted on the display device 20, making it easier for the user to select parameters for manufacturing the product so as to obtain desired output information. In this way, the data analysis support device 10 can support the user in inputting input information.

[0095] The objective variables may define a plurality of parameters indicating the properties of the product, and the distribution map data may include objective variable graphs 231 and 232, which are an example of information indicating the relationship between the plurality of objective variable parameters. With this configuration, the user can grasp the distribution of the objective variables in the data set 121 by looking at the objective variable graphs 231 and 232. The user can accurately predict the output information resulting from various selections of input information by checking where the output information obtained using the prediction model 122 is located on the objective variable graphs 231 and 232.

[0096] The control unit 11 may generate histograms 233, 234 of the objective variables, and display the histograms 233, 234 together with the distribution diagram on the display device 20. According to this configuration, the user can grasp the reliability of the output information obtained using the prediction model 122 by looking at the histograms 233, 234 together with the objective variable graphs 231, 232.

[0097] The prediction model 122 may be a trained model trained to output a response variable in response to an input of an explanatory variable. In this case, the control unit 11 inputs the input information to the prediction model 122, thereby acquiring output information output from the prediction model 122.

[0098] The control unit 11 may determine trend information indicating the trend of change in the objective variable when the explanatory variable changes based on the prediction model 122, and may superimpose a trend line, which is an example of an image indicated by the trend information, on the distribution map. By checking the trend line, the user can understand how the objective variable changes when the explanatory variable is changed.

[0099] The control unit 11 may cause the display device 20 to display reference information indicating at least one of the range of the explanatory variables and the range of the objective variables of the data set 121. According to this configuration, the user can determine input information by referring to the reference information.

[0100] The reference information may include information indicating a range of the explanatory variable. The control unit 11 may input reference input information within the range of the explanatory variable of the reference information to the prediction model, thereby acquiring reference output information output from the prediction model, and display the reference input information and the reference output information on the display device 20. According to this configuration, the user can efficiently select input information to obtain desired output information by referring to the reference input information and the reference output information.

[0101] 5. Other embodiments As described above, the embodiment has been described as an example of the technology of the present invention. However, the technology of the present invention is not limited to this, and can be applied to embodiments in which modifications, substitutions, additions, omissions, etc. are appropriately performed. In addition, it is also possible to combine the components described in the above embodiment to create a new embodiment. Therefore, other embodiments will be exemplified below.

[0102] In the above embodiment, an example has been described in which the data set 121 and the prediction model 122 are stored in the storage unit 12 in advance, but the present invention is not limited to this. For example, the control unit 11 may import an external data set created externally via the input interface 13. The control unit 11 may also import an external model created externally. The control unit 11 can perform operations such as prediction and search using an external data set and / or an external model instead of or in addition to the data set 121 and the prediction model 122.

[0103] Such an external model has the same class as the prediction model 122 already installed in the data analysis support device 10 or a compatible class. The class is a template that defines the format, method, etc. of input data and output data. The control unit 11 is configured to accept only an external model created by inheriting the same class as the prediction model 122 or a compatible class. In this case, the control unit 11 compares the input data format, output data format, and method of the prediction model 122 and the external model, and if they match, determines that the classes of the prediction model 122 and the external model are the same. For example, a model of the same class as the prediction model 122 refers to a model that can derive a target variable Y as an output by a calculation expressed as Y=f(X) using an explanatory variable X as an input. When the control unit 11 receives a model that is not the same class as the prediction model 122 or a compatible class, it does not import the model and displays an error message on the display device 20.

[0104] This allows the types of prediction model 122, external model, and data set already installed in the data analysis support device 10 to be unified, and the control unit 11 can treat the prediction model 122 and the external model in the same way when using them for prediction and exploration.

[0105] The prediction model 122 and the external model are not limited to trained models. For example, the prediction model 122 and the external model may be a model that does not require training, such as a model that calculates Y expressed as Y=aX (a is a constant).

[0106] 16 is a flowchart illustrating an example of an external information import process according to another embodiment. The external information includes at least one of an external dataset and an external model.

[0107] In FIG. 16, the control unit 11 acquires an external data set or an external model via the input interface 13 (S31).

[0108] Next, the control unit 11 determines whether or not there is a data set or a model in the storage unit 12 that matches the class information of the external data set or the external model (S32).

[0109] If such a data set or model is present in the storage unit 12 (Yes in S32), the control unit 11 stores the external data set or external model acquired in step S31 in the storage unit 12 (S33). In this case, the control unit 11 also stores in the storage unit 12 identification information (ID information) of the stored external data set or external model.

[0110] If there is no dataset or model in the memory unit 12 that matches the class information of the external dataset or external model (No in S32), the control unit 11 does not import (save) the dataset or model acquired in step S31, and displays an error message on the display device 20.

[0111] 6.Examples The following are examples of aspects of the present invention.

[0112] <Aspect 1> A storage unit that stores a data set including explanatory variables that define at least one parameter for manufacturing a product and a response variable that defines at least one parameter that indicates a property of the product, and a prediction model that associates the explanatory variables with the response variable; an input unit that receives input information corresponding to the explanatory variables; a control unit that obtains output information indicating a property of the product using the prediction model based on the input information, The control unit is generating distribution map data showing a relationship between the explanatory variables and the objective variable of the data set; Plotting the input information and the output information on a distribution map indicated by the distribution map data; displaying the distribution map on which the input information and the output information are plotted on a display device; Data analysis support device.

[0113] <Aspect 2> The objective variables define a plurality of parameters indicative of properties of the product; the distribution map data includes information indicating a relationship between the plurality of parameters of the response variable, 2. A data analysis support device according to claim 1.

[0114] <Aspect 3> The control unit is generating a histogram of the response variable; displaying the histogram together with the distribution map on the display device; 3. The data analysis support device according to claim 1 or 2.

[0115] <Aspect 4> The prediction model is a trained model trained to output the objective variable in response to an input of the explanatory variable, The control unit inputs the input information to the prediction model, and thereby acquires the output information output from the prediction model. The data analysis support device according to any one of aspects 1 to 3.

[0116] <Aspect 5> The control unit is determining trend information indicative of a trend of change in the objective variable when the explanatory variable changes based on the prediction model; an image indicating the trend information is superimposed on the distribution map; The data analysis support device according to any one of aspects 1 to 4.

[0117] <Aspect 6> 2. The data analysis support device according to aspect 1, wherein the control unit causes the display device to display reference information indicating at least one of a range of the explanatory variables and a range of the target variable of the data set.

[0118] <Aspect 7> the reference information includes information indicating a range of the explanatory variable, The control unit is inputting reference input information within the range of the explanatory variables of the reference information into the prediction model, thereby obtaining reference output information output from the prediction model; 7. The data analysis support device according to claim 6, wherein the reference input information and the reference output information are displayed on the display device.

[0119] <Aspect 8> The control unit is an external data set including the explanatory variables and the objective variable, and an external model associating the explanatory variables with the objective variable, are externally acquired via the input unit; When the external data set or the external model is acquired via the input unit, determining whether class information of the acquired external data set or the external model matches class information of the data set or the prediction model; When it is determined that the class information of the external dataset or the external model matches the class information of the dataset or the prediction model, the external dataset or the external model is stored in the storage unit. The data analysis support device according to any one of aspects 1 to 7.

[0120] <Aspect 9> A data analysis support method executed by a control unit, A step of preparing a data set including explanatory variables defining at least one parameter for manufacturing a product and a response variable defining at least one parameter indicating a property of the product, and a prediction model relating the explanatory variables and the response variable; acquiring input information corresponding to the explanatory variables; using the predictive model to obtain output information indicative of a property of the product based on the input information; generating distribution map data showing a relationship between the explanatory variables and the objective variable of the data set; Plotting the input information and the output information on a distribution map represented by the distribution map data; and displaying, on a display device, the distribution map in which the input information and the output information are plotted. Methods to support data analysis. [Explanation of symbols]

[0121] 1. Data analysis support system 10 Data analysis support device 11 Control section 12 Storage section 13 Input Interface 14 Output Interface 20 Display device 21 Model selection screen 23 First graph screen 24 Second Graph Screen 24a 3rd Graph Screen 25 Search settings screen 26 Search results screen 111 Arithmetic section 112 Graph Generation Unit 121 Datasets 122 Predictive Models 322a Prediction execution screen 322b Prediction execution screen

Claims

1. a storage unit that stores a data set including explanatory variables that define at least one parameter for manufacturing a product and a response variable that defines at least one parameter that indicates a property of the product, and a prediction model that associates the explanatory variables with the response variable; an input unit that receives input information corresponding to the explanatory variables; a control unit that obtains output information indicating a property of the product using the prediction model based on the input information, The control unit is generating distribution map data showing a relationship between the explanatory variables and the objective variable of the data set; Plotting the input information and the output information on a distribution map indicated by the distribution map data; displaying the distribution map on which the input information and the output information are plotted on a display device; Data analysis support device.

2. The objective variables define a plurality of parameters indicative of properties of the product; the distribution map data includes information indicating a relationship between the plurality of parameters of the response variable, The data analysis support device according to claim 1 .

3. The control unit is generating a histogram of the response variable; displaying the histogram together with the distribution map on the display device; The data analysis support device according to claim 1 .

4. The prediction model is a trained model trained to output the objective variable in response to an input of the explanatory variable, The control unit inputs the input information to the prediction model, and thereby acquires the output information output from the prediction model. The data analysis support device according to claim 1 .

5. The control unit is determining trend information indicative of a trend of change in the objective variable when the explanatory variable changes based on the prediction model; an image indicating the trend information is superimposed on the distribution map; The data analysis support device according to claim 1 .

6. The data analysis support device according to claim 1 , wherein the control unit causes the display device to display reference information indicating at least one of a range of the explanatory variables and a range of the response variable of the data set.

7. the reference information includes information indicating a range of the explanatory variable, The control unit is inputting reference input information within the range of the explanatory variables of the reference information into the prediction model, thereby obtaining reference output information output from the prediction model; The data analysis support device according to claim 6 , wherein the reference input information and the reference output information are displayed on the display device.

8. The control unit is an external data set including the explanatory variables and the objective variable, and an external model associating the explanatory variables with the objective variable, are externally acquired via the input unit; When the external data set or the external model is acquired via the input unit, determining whether class information of the acquired external data set or the external model matches class information of the data set or the prediction model; When it is determined that the class information of the external dataset or the external model matches the class information of the dataset or the prediction model, the external dataset or the external model is stored in the storage unit. The data analysis support device according to any one of claims 1 to 7.

9. A data analysis support method executed by a control unit, A step of preparing a data set including explanatory variables defining at least one parameter for manufacturing a product and a response variable defining at least one parameter indicating a property of the product, and a prediction model relating the explanatory variables and the response variable; acquiring input information corresponding to the explanatory variables; using the predictive model to obtain output information indicative of a property of the product based on the input information; generating distribution map data showing a relationship between the explanatory variables and the objective variable of the data set; Plotting the input information and the output information on a distribution map represented by the distribution map data; and displaying, on a display device, the distribution map in which the input information and the output information are plotted. Methods to support data analysis.

Images