JP2021179668A - Data analysis system, data analysis method, and data analysis program - Google Patents

Abstract

To provide a data analysis system, a data analysis method, and a data analysis program capable of facilitating data analysis.SOLUTION: In a data analysis system, at least one processor receives a data set containing a plurality of data units being a collection of pieces of data of a plurality of items, creates, on the basis of the data set, a plurality of respectively different regression models in which one item among the plurality of items is an objective variable and another item among the plurality of items is an explanatory variable, calculates accuracy of each of the plurality of regression models, and causes a display unit to display, as a display model, a regression model selected from among the plurality of regression models according to the accuracy. Each of the plurality of regression models is a regression model using a regression formula or a regression model using a decision tree.SELECTED DRAWING: Figure 3

Description

The present invention relates to a data analysis system, a data analysis method and a data analysis program.

As a data analysis method, a method is known in which a regression model representing the relationship between the objective variable and the explanatory variable is created based on the data set to be analyzed, and analysis is performed based on the created regression model (for example). See Patent Document 1).

Japanese Unexamined Patent Publication No. 2020-24544

Various regression models can be used in the data analysis method as described above, but since the suitable regression model differs depending on the data set, it is necessary for the user to select an appropriate regression model, and data analysis is not easy. There is. In addition, some regression models are not easy to interpret, which can also make data analysis difficult.

An object of the present invention is to provide a data analysis system, a data analysis method, and a data analysis program capable of facilitating data analysis.

The data analysis system of the present invention includes at least one processor, and the at least one processor accepts a data set containing a plurality of data units which are a collection of data of a plurality of items, and based on the data set, a plurality of items. One of the items is the objective variable, and the other item of the multiple items is the explanatory variable. Create multiple regression models that are different from each other, calculate the accuracy of each of the multiple regression models, and multiple regressions. A regression model selected from the models according to the accuracy is displayed as a display model on the display unit, and each of the plurality of regression models is a regression model using a regression equation or a regression model using a decision tree. Is.

In this data analysis system, a plurality of different regression models are generated based on the data set, and the accuracy of each created regression model is calculated. Then, the regression model selected from the plurality of regression models according to the accuracy is displayed on the display unit. This allows the user to perform analysis using the regression model selected according to the accuracy. As a result, data analysis can be easily performed. Further, in this data analysis system, each regression model is a regression model using a regression equation or a regression model using a decision tree. This makes it easy to interpret the regression model using the regression equation and the regression model using the decision tree, so that the user can easily interpret the displayed regression model. Therefore, according to this data analysis system, data analysis can be facilitated.

At least one processor may display the regression model with the highest accuracy among the plurality of regression models on the display unit as a display model. In this case, data analysis can be further facilitated.

At least one processor may display a graph showing the relationship between the value of the objective variable in the data set and the value of the objective variable predicted by the display model on the display unit together with the display model. In this case, data analysis can be further facilitated.

At least one processor may display the importance of the explanatory variables in the display model on the display unit together with the display model. In this case, data analysis can be further facilitated.

At least one processor may display the tuning result of the hyperparameters in the display model on the display unit together with the display model. In this case, data analysis can be further facilitated.

At least one processor may display a plurality of labels indicating each of the plurality of regression models on the display unit so that the corresponding regression models are arranged in descending order of accuracy. In this case, the user can easily compare the accuracy of the plurality of regression models, and the data analysis can be further facilitated.

At least one processor calculates the accuracy of each of the plurality of regression models using a plurality of accuracy indexes, and displays the accuracy of the plurality of regression models when each of the plurality of accuracy indexes is used on the display unit. May be good. In this case, the user can easily compare the accuracy of the plurality of regression models when each accuracy index is used, and the data analysis can be further facilitated.

At least one processor may display the coefficients of the explanatory variables in each of the plurality of regression models on the display unit. In this case, data analysis can be further facilitated.

At least one processor may display the importance of the explanatory variables in each of the plurality of regression models on the display unit. In this case, data analysis can be further facilitated.

The data analysis method of the present invention is a data analysis method executed by a data analysis system including at least one processor, and is a step of accepting a data set including a plurality of data units which are a set of data of a plurality of items, and data. Based on the set, the step of creating multiple different regression models with one of the multiple items as the objective variable and the other item of the multiple items as the explanatory variable, and the multiple regression models. Each of the plurality of regression models is provided with a step of calculating the accuracy of each and a step of displaying a regression model selected according to the accuracy from a plurality of regression models on the display unit as a display model, and each of the plurality of regression models is a regression equation. Is a regression model using, or is a regression model using a decision tree. According to this data analysis method, data analysis can be facilitated for the reasons described above.

The data analysis program of the present invention has a step of accepting a data set including a plurality of data units which are a set of data of a plurality of items, and a plurality of items with one of the plurality of items as an objective variable based on the data set. A step to create multiple regression models different from each other using other items of the items as explanatory variables, a step to calculate the accuracy of each of the multiple regression models, and a step to calculate the accuracy of each of the multiple regression models according to the accuracy. A computer is made to execute a step of displaying the selected regression model as a display model on the display unit, and each of the plurality of regression models is a regression model using a regression equation or a regression model using a decision tree. Is. According to this data analysis program, data analysis can be facilitated for the reasons described above.

According to the present invention, it is possible to provide a data analysis system, a data analysis method, and a data analysis program that can facilitate data analysis.

It is a figure which shows the example of the functional structure of the data analysis system which concerns on embodiment. It is a figure which shows the configuration example of the hardware of the computer which constitutes a data analysis system. It is a flowchart which shows the operation example of a data analysis system. It is a figure which shows the example of a data set. It is a figure which shows the display example of the display part. It is a figure which shows the display example of the regression model using a decision tree. (A) is a diagram showing a state before the selection box for selecting a display model is expanded, and (b) is a diagram showing a state in which the selection box is expanded. It is a figure which shows the example of the display for comparison of accuracy. It is a figure which shows the example of the display for comparison of the coefficient of the explanatory variable. It is a figure which shows the example of the display for comparison of the importance of an explanatory variable.

Hereinafter, an embodiment of the present invention will be described in detail with reference to the drawings. In the following description, the same reference numerals will be used for the same or equivalent elements, and duplicate description will be omitted.
[System configuration]

As shown in FIG. 1, the data analysis system 1 according to the embodiment includes a reception unit 11, a model creation unit 12, an accuracy calculation unit 13, and a display control unit 14 as functional elements. The reception unit 11 receives the data set 30. The model creation unit 12 creates a plurality of regression models based on the data set 30. The accuracy calculation unit 13 calculates the accuracy of each regression model. The display control unit 14 causes the display unit 26, which will be described later, to display the regression model selected from the plurality of regression models.

The data analysis system 1 is composed of, for example, a computer 20. As shown in FIG. 2, the computer 20 includes a processor 21, a main storage unit 22, an auxiliary storage unit 23, a communication control unit 24, an input unit 25, and a display unit 26. The processor 21 is, for example, a CPU, and executes an operating system, an application program, and the like. The main storage unit 22 is composed of, for example, a ROM, a RAM, or the like. The auxiliary storage unit 23 is composed of, for example, a hard disk, a flash memory, or the like, and stores a larger amount of data than the main storage unit 22. The communication control unit 24 is composed of, for example, a network card, a wireless communication module, or the like. The input unit 25 is composed of, for example, a keyboard, a mouse, a touch panel, and the like. The display unit 26 is composed of, for example, a monitor, a touch panel display, or the like.

Each functional element of the data analysis system 1 is realized by executing a data analysis program 27 stored in advance in the auxiliary storage unit 23. Specifically, by loading the data analysis program 27 on the processor 21 or the main storage unit 22 and causing the processor 21 to execute the data analysis program 27, the reception unit 11, the model creation unit 12, the accuracy calculation unit 13, and the accuracy calculation unit 13 Each function of the display control unit 14 is realized. The processor 21 operates the communication control unit 24, the input unit 25, and the display unit 26 according to the data analysis program 27, and reads and writes data in the main storage unit 22 and the auxiliary storage unit 23. The data or database required for processing is stored in the main storage unit 22 or the auxiliary storage unit 23.

The data analysis program 27 may be provided after being fixedly recorded on a tangible recording medium such as a CD-ROM, a DVD-ROM, or a semiconductor memory. Alternatively, the data analysis program 27 may be provided via a communication network as a data signal superimposed on a carrier wave.

The data analysis system 1 may be configured by one computer 20 or may be configured by a plurality of computers 20. When a plurality of computers 20 are used, one data analysis system 1 may be logically constructed by connecting these computers 20 to each other via a communication network such as the Internet or an intranet.
[System operation]

An example of a data analysis method executed by the data analysis system 1 will be described with reference to FIG. First, the reception unit 11 receives the data set 30 (step S1). The input of the data set 30 to the reception unit 11 is performed by, for example, by the user via the input unit 25 and the display unit 26. For example, when the user specifies the data set 30 stored in the auxiliary storage unit 23, the designated data set 30 is read and accepted by the reception unit 11.

The data set 30 includes a plurality of data units 31 which are a set of data of a plurality of items. The items included in the data unit 31 may be arbitrarily set. The item may be, for example, the characteristics, composition, or the like of a material, a compound, or the like, or may be the characteristics, dimensions, a material, or the like of an apparatus, a device, or the like. The data of the item may be a numerical value or a character. Character data is converted into numerical data and used. Some items may have no data (missing values). The number of data units 31 is not limited, but may be, for example, several hundred or less.

FIG. 4 is a diagram showing an example of the data set 30. In this example, the dataset 30 is represented in tabular form. Each row corresponds to the data unit 31, and each column corresponds to an item. Each data unit 31 represents a relationship between material composition and magnetic permeability. The data unit 31 includes, as items, magnetic permeability, material A, material B, material C, material D, material E, material F, and material H. The magnetic permeability data is a numerical value indicating the magnetic permeability, and the data of the materials A to H is a numerical value expressing the material composition as a percentage.

Following step S1, the reception unit 11 receives the analysis condition (step S2). The input of the analysis condition to the reception unit 11 is performed by, for example, by the user via the input unit 25 and the display unit 26. The analysis conditions include the specification of the objective variable and the design variable. The user selects one item from the plurality of item variables in the data set 30 as the objective variable, and selects one or more items from the remaining items as the explanatory variables. For example, a selection box for selecting an objective variable is displayed on the display unit 26, and the user selects the objective variable by selecting an item in the selection box. Further, a plurality of check boxes corresponding to items other than the item selected as the objective variable are displayed on the display unit 26, and the user selects an explanatory variable by checking the corresponding check boxes.

In addition, the analysis conditions include the specification of the missing value processing method. The user selects one missing value processing method used in the model creation process (step S3) described later from the plurality of missing value processing methods. For example, the display unit 26 displays a selection box for selecting a missing value processing method, and by selecting the missing value processing method in the selection box, the user selects the missing value processing method to be used. Examples of the missing value processing method include, for example, Listwise deletion, Predictive meanmatching, and random sampling. In listwise removal, the data unit 31 containing the missing value is not analyzed. As the missing value processing method, a method of treating the missing value as a value 0 (zero substitution) or no missing value processing (none) may be selectable.

In addition, the analysis conditions include the designation of the interaction term and the square term. Setting the interaction term to Yes adds an interaction term between the two explanatory variables to the regression model. If the squared term is set to Yes, the squared term of the explanatory variable is added to the regression model. The user chooses whether to add the interaction and squared terms to the regression model. For example, the display unit 26 displays a group of check boxes arranged in a matrix so that each of the row direction and the column direction corresponds to an explanatory variable, and the user can add by checking the corresponding check boxes. Select the interaction term and the square term. Below the check box group, for example, the importance of each explanatory variable calculated by gradient boosting may be displayed. In this case, the user can select the interaction term and the square term while referring to the importance.

Further, the analysis condition may include a specification of whether to use the objective variable as it is or to convert it into a logarithm and use it. Further, the analysis condition may include the specification of the numerical range for each explanatory variable. Further, the analysis condition may include designation of which data unit 31 is to be analyzed.

In addition, the analysis conditions include the specification of the regression model. The user selects a plurality of regression models to be used in the model creation process (step S3) described later from a plurality of regression models different from each other. For example, the display unit 26 displays a selection box for selecting a regression model, and by selecting the regression model in the selection box, the user selects the regression model to be used. Hereinafter, the regression model displayed on the display unit 26 and selectable by the user is also referred to as a selectable regression model.

The selectable regression model is a regression model using a regression equation or a regression model using a decision tree. Selectable regression models include, for example, Bayesian Generalized Linear Model (Bayesian Generalized Linear Model), GeneralizedLinear Model (GeneralizedLinear Model), and Multivariate Adaptive Regression Spline (MARS) as regression models using regression equations. ), Negative BinomialGeneralized Linear Model, Partial LeastSquares, Linear Regression, Principal Component Regression, PenalizedLinear Regression), Elastic Net, LASSO, Ridge, etc. are included. The selectable regression model includes, for example, a conditional inference tree, a Cubist, a CART, and the like as a regression model using a decision tree.

Selectable regression models do not include neural networks and support vector machines, as well as regression models using ensemble learning of nonlinear models. These regression models are black boxes and are not easy to interpret. Ensemble learning is a method of creating a plurality of regression models and combining the results to create one regression model. Non-linear model ensemble learning refers to those that include a non-linear model as a regression model to be combined. The regression model using the regression equation is, for example, a regression model using a regression equation expressed by a combination of a monomial expression or a polynomial of order n (n is an integer of 1 or more) or a fractional expression including them. The regression equation may be composed of only the intercept (constant term). The regression equation may include an exponential function, but those including an exponential function (double exponential function) may be excluded as variables of the exponential function. The regression equation may include a max function, but those containing a max function may be excluded. The regression equation may include an exponential function and a max function, but those containing the max function as a variable of the exponential function, those containing the exponential function as the variable of the max function, and those containing the max function as the variable of the max function It may be excluded. Those containing spline functions may be excluded from the regression equation.

In addition, the analysis conditions include the specification of the tuning range of the hyperparameters of each regression model. For example, if the user selects a large numerical value as an index indicating the tuning range, the tuning range of the hyperparameter becomes large. On the other hand, if the user selects a small numerical value as an index indicating the degree of tuning, the tuning range of the hyperparameter becomes small. The hyperparameter is a parameter indicating the degree to which the regularization term is considered, for example, when the regression model is LASSO, Ridge or ElasticNet. Hyperparameters are parameters that represent a link function, for example, if the regression model is a negative binomial distribution generalized linear model. For example, when the regression model uses a decision tree, the hyperparameter may be a parameter representing the maximum value of the depth of the tree or a parameter representing the number of terminal nodes. If the regression model is CART, the hyperparameters are complex parameters.

In addition, the analysis conditions include the specification of the data verification method. The user selects one data verification method used in the model creation process (step S3) described later from the plurality of data verification methods. For example, the display unit 26 displays a selection box for selecting a data verification method, and by selecting the data verification method in the selection box, the user selects the data verification method to be used. Examples of data validation methods include, for example, k-fold cross-validation, bootstrap, leave-one-out cross-validation, and the like.

Following step S2, the model creation unit 12 creates a plurality of regression models (step S3). More specifically, the model creation unit 12 creates a plurality of regression models selected in step S2 using the analysis conditions specified in step S2 based on the data set 30 received in step S1. ..

For example, when k-partition cross-validation is selected as the data validation method, the procedure for creating each regression model is as follows. First, tune the hyperparameters. Specifically, a plurality of data units 31 are randomly divided into k (k is an integer of 2 or more). The hyperparameters are changed within the tuning range set as the analysis condition, and k-fold cross-validation is performed to compare the accuracy. For example, when the objective variable is a numerical value, the accuracy is compared by RMSE (root mean square error), and when the objective variable is a character, the accuracy is compared by the correct answer rate (Accuracy). Select the value of the hyperparameter with the highest accuracy. If the regression model is an algorithm that does not require hyperparameter tuning, hyperparameter tuning is not performed.

Then, the accuracy is calculated using the selected hyperparameters, and it is confirmed whether or not the expected accuracy is obtained. Specifically, the data is divided into k again (however, the division is different from that at the time of tuning the hyperparameters), and the accuracy is calculated and confirmed using the selected hyperparameters. Subsequently, each regression model based on all the data units 31 is created using the selected hyperparameters.

Following step S3, the accuracy calculation unit 13 calculates the accuracy of each regression model (step S4). In this example, the accuracy calculation unit 13 calculates the accuracy of each regression model using a plurality of accuracy indexes. Examples of the accuracy index used when the objective variable is a numerical value include RMSE, coefficient of determination (R-squared), MAE (mean absolute error) and the like. Examples of the accuracy index used when the objective variable is a character include a correct answer rate, a precision rate, a recall rate, and the like.

Following step S4, the display control unit 14 causes the display unit 26 to display the regression model with the highest accuracy as a display model (step S5). In this example, the display control unit 14 causes the display unit 26 to display the regression model having the smallest RMSE.

FIG. 5 is a diagram showing a display example of the display unit 26. This example is an example in which the display model 41 is a negative binomial distribution generalized linear model, and the regression equation of the display model 41 is displayed on the display unit 26. The objective variable is magnetic permeability and the explanatory variables are materials A to H. No interaction term and square term have been added. The variable P1 is an intercept, and the variables P2, P3, P4, P5, P6, and P7 are coefficients of materials A, B, E, F, G, and H, which are explanatory variables, respectively. In this example, the other explanatory variables C and D are not included in the regression equation. The display unit 26 displays the graph 42, the graph 43, the variable importance 44, and the details 45 of the analysis result together with the display model 41.

The graph 42 is displayed above the display model 41. FIG. 42 is a scatter diagram showing the relationship between the value of the objective variable (measured value) in the data set 30 and the value of the objective variable (predicted value) predicted by the display model 41. In the graph 42, the horizontal axis is the measured value and the vertical axis is the predicted value. In the graph 42, a straight line indicating the plot position when the measured value and the predicted value are equal is displayed by a broken line. In the graph 42, the values of RMSE and the coefficient of determination are displayed.

The graph 43 is displayed above the display model 41. The graph 43 is a line graph showing the tuning result of the hyperparameters in the display model 41. In graph 43, the horizontal axis is hyperparameters and the vertical axis is RMSE. From the graph 43, it can be seen that in this example, when the hyperparameter is B, the RMSE is the smallest, and the hyperparameter B is selected from the hyperparameters A to C.

The variable importance 44 is displayed above the display model 41. Graph 42, graph 43, and variable importance 44 are arranged in this order in the left-right direction. The variable importance 44 is a display indicating the importance of each explanatory variable in the display model 41 numerically. In this example, the explanatory variables A to H are displayed so as to be arranged from the top in descending order of importance. The details 45 of the analysis result are displayed below the display model 41. Details 45 of the analysis result show various information about the display model 41.

FIG. 6 is a diagram showing a display example when the display model is a regression model using a decision tree. This example is an example in which the display model 41 is CART, and the decision tree of the display model 41 is displayed. In this example, the decision tree branches according to the magnitude of the items A and C, which are explanatory variables. The display other than the display model 41 is the same as in FIG. 5, for example. When the regression model using the decision tree is the display model 41, the conditional branch of the display model 41 may be displayed in place of or in addition to the decision tree of the display model 41.

Further, the display control unit 14 may display the regression model selected by the user on the display unit 26 as the display model 41. For example, in the example of FIG. 7, the display unit 26 displays a selection box 51 for selecting the display model 41. The selection box 51 is displayed together with, for example, the display model 41, but may be displayed on a screen (tab) different from that of the display model 41.

As shown in FIG. 7A, in the state before expansion, the selection box 51 displays a label indicating the currently selected display model 41. In this example, the currently selected display model 41 is a regression model D, and its name is displayed as a label indicating the display model 41. The label is not limited to the name, and may be a character, a symbol, a figure, or the like indicating a regression model.

As shown in FIG. 7B, when the user presses the selection box 51, the selection box 51 is expanded. In the expanded state, labels indicating a plurality of regression models for which models have been created are displayed in the selection box 51 so as to be arranged from the top in descending order of accuracy of the corresponding regression models. The user selects the regression model to be displayed as the display model 41 by selecting the label indicating the regression model in the selection box 51. Upon receiving this selection, the display control unit 14 displays the selected regression model as the display model 41 on the display unit 26.

Further, the display control unit 14 may display the accuracy of each regression model when each accuracy index is used on the display unit 26. For example, in the example of FIG. 8, the display unit 26 displays a table 52 showing the accuracy of each regression model when each accuracy index is used. Table 52 is displayed on a screen (tab) different from that of the display model 41, for example, but may be displayed together with the display model 41 and the like. In Table 52, the regression models A to D are displayed so that the RMSEs are arranged from the top in ascending order. The user can sort the regression models A to D in descending order of the accuracy index by pressing the label of the accuracy index.

Further, the display control unit 14 may display the coefficients of each explanatory variable in each regression model on the display unit 26. For example, in the example of FIG. 9, the display unit 26 displays a table 53 showing the coefficients and intercept values of the explanatory variables (materials A to H) in the regression models A to D. Table 53 is displayed on a screen (tab) different from that of the display model 41, for example, but may be displayed together with the display model 41 and the like.

Further, the display control unit 14 may display the importance of each explanatory variable in each regression model on the display unit 26. For example, in the example of FIG. 10, the display unit 26 displays a table 54 showing the importance of each explanatory variable (materials A to H) in each regression model. Table 54 is displayed on a screen (tab) different from that of the display model 41, for example, but may be displayed together with the display model 41 and the like. In Table 54, the explanatory variables are displayed so as to be arranged from the top in descending order of importance in the regression model A. By pressing the label of the regression model, the user can sort the explanatory variables in descending order of importance in the regression model.

Further, the data analysis system 1 is configured to be able to receive the data set 30 for prediction by the reception unit 11. The user inputs the data set 30 for prediction to the reception unit 11 via the input unit 25 and the display unit 26. When the reception unit 11 receives the data set 30 for prediction, the display control unit 14 displays the prediction result of the explanatory variable by the current display model 41 on the display unit 26.
[Action and effect]

In the data analysis system 1, a plurality of regression models different from each other are generated based on the data set 30, and the accuracy of each created regression model is calculated. Then, the regression model selected from the plurality of regression models according to the accuracy is displayed on the display unit 26. This allows the user to perform analysis using the regression model selected according to the accuracy. As a result, data analysis can be easily performed. Further, in the data analysis system 1, each regression model is a regression model using a regression equation or a regression model using a decision tree. This makes it easy to interpret the regression model using the regression equation and the regression model using the decision tree, so that the user can easily interpret the displayed regression model. Therefore, according to the data analysis system 1, data analysis can be facilitated. As a result, it becomes possible to easily perform a quantitative explanation and a highly accurate prediction about the data set 30.

The processor 21 causes the display unit 26 to display the regression model with the highest accuracy among the plurality of regression models as the display model 41. This makes it possible to further facilitate data analysis.

The processor 21 causes the display unit 26 to display a graph 42 showing the relationship between the value of the objective variable in the data set 30 and the value of the objective variable predicted by the display model 41 together with the display model 41. This makes it possible to further facilitate data analysis.

The processor 21 causes the display unit 26 to display the importance of the explanatory variables (variable importance 44) in the display model 41 together with the display model 41. This makes it possible to further facilitate data analysis.

The processor 21 causes the display unit 26 to display the tuning result (graph 43) of the hyperparameters in the display model 41 together with the display model 41. This makes it possible to further facilitate data analysis.

The processor 21 causes the display unit 26 to display a plurality of labels indicating each of the plurality of regression models so that the corresponding regression models are arranged in descending order of accuracy (FIG. 7 (b)). As a result, the accuracy of a plurality of regression models can be easily compared by the user, and data analysis can be further facilitated.

The processor 21 causes the display unit 26 to display the accuracy of the plurality of regression models when each of the plurality of accuracy indexes is used (FIG. 8). As a result, the user can easily compare the accuracy of a plurality of regression models when each accuracy index is used, and data analysis can be further facilitated.

The processor 21 causes the display unit 26 to display the coefficients of the explanatory variables in each of the plurality of regression models (FIG. 9). This makes it possible to further facilitate data analysis.

The processor 21 causes the display unit 26 to display the importance of the explanatory variables in each of the plurality of regression models (FIG. 10). This makes it possible to further facilitate data analysis.

The present invention is not limited to the above embodiment. For example, in step S5 of the above embodiment, the regression model having the highest accuracy among the plurality of regression models is displayed on the display unit 26 as the display model 41, but the process in step S5 is not limited to this, and is the next process. You may. Labels indicating multiple regression models that have been modeled are displayed so as to be arranged from the top in descending order of accuracy of the corresponding regression models. The user selects a regression model to be displayed as the display model 41 by selecting one label. Upon receiving this selection, the display control unit 14 displays the selected regression model as the display model 41 on the display unit 26.

1 ... Data analysis system, 20 ... Computer, 21 ... Processor, 26 ... Display unit, 27 ... Data analysis program, 30 ... Data set, 31 ... Data unit, 41 ... Display model, 42 ... Graph.

Claims (11)

Equipped with at least one processor
The at least one processor
Accepts a dataset that contains multiple data units that are a collection of data for multiple items,
Based on the data set, we created a plurality of different regression models in which one of the plurality of items is the objective variable and the other item of the plurality of items is the explanatory variable.
The accuracy of each of the plurality of regression models was calculated, and
The regression model selected from the plurality of regression models according to the accuracy is displayed on the display unit as a display model.
A data analysis system in which each of the plurality of regression models is a regression model using a regression equation or a regression model using a decision tree. The at least one processor
The data analysis system according to claim 1, wherein the regression model having the highest accuracy among the plurality of regression models is displayed on the display unit as the display model. The at least one processor
The invention according to claim 1 or 2, wherein a graph showing a relationship between the value of the objective variable in the data set and the value of the objective variable predicted by the display model is displayed on the display unit together with the display model. Data analysis system. The at least one processor
The data analysis system according to any one of claims 1 to 3, wherein the importance of the explanatory variables in the display model is displayed on the display unit together with the display model. The at least one processor
The data analysis system according to any one of claims 1 to 4, wherein the tuning result of the hyperparameter in the display model is displayed on the display unit together with the display model. The at least one processor
The data analysis according to any one of claims 1 to 5, wherein a plurality of labels indicating the plurality of regression models are displayed on the display unit so as to arrange the plurality of labels indicating the plurality of regression models in descending order of the accuracy of the corresponding regression models. system. The at least one processor
For each of the plurality of regression models, the accuracy is calculated using a plurality of accuracy indexes.
The data analysis system according to any one of claims 1 to 6, wherein the accuracy of the plurality of regression models when each of the plurality of accuracy indexes is used is displayed on the display unit. The at least one processor
The data analysis system according to any one of claims 1 to 7, wherein the coefficient of the explanatory variable in each of the plurality of regression models is displayed on the display unit. The at least one processor
The data analysis system according to any one of claims 1 to 8, wherein the importance of the explanatory variables in each of the plurality of regression models is displayed on the display unit. A data analysis method performed by a data analysis system comprising at least one processor.
A step that accepts a dataset that contains multiple data units that are a set of data for multiple items,
Based on the data set, a step of creating a plurality of different regression models with one of the plurality of items as the objective variable and the other item of the plurality of items as the explanatory variables, and
Steps to calculate the accuracy of each of the plurality of regression models,
A step of displaying the regression model selected from the plurality of regression models according to the accuracy as a display model on the display unit is provided.
A data analysis method in which each of the plurality of regression models is a regression model using a regression equation or a regression model using a decision tree. A step that accepts a dataset that contains multiple data units that are a set of data for multiple items,
Based on the data set, a step of creating a plurality of different regression models with one of the plurality of items as the objective variable and the other item of the plurality of items as the explanatory variables, and
Steps to calculate the accuracy of each of the plurality of regression models,
A computer is made to execute a step of displaying the regression model selected from the plurality of regression models according to the accuracy on the display unit as a display model.
A data analysis program in which each of the plurality of regression models is a regression model using a regression equation or a regression model using a decision tree.

Images