JP7292235B2 - Analysis support device and analysis support method - Google Patents

Description

The present invention relates to an analysis support device and an analysis support method.

In recent years, it is possible to collect business data in many industrial fields, and there is a need to support and automate data analysis such as generating models that predict KPIs (Key Performance Indicators) in corporate activities such as sales, and factor analysis. is rising. In addition, tools have been developed that learn from past analysis cases and support preprocessing, feature generation, and modeling required for data analysis.

As background art in this technical field, there is International Publication No. 2019/073900 (Patent Document 1). In this publication, "Disease can be easily and highly accurately determined based on body sounds caused by body activity. Learning a determination algorithm D2 for determining disease based on body sounds caused by body activity A machine learning device (3), comprising: a sound information acquisition unit (33) for acquiring sound information of the body sounds of a plurality of subjects; and a diagnostic information acquisition unit for acquiring diagnostic information regarding diseases of the plurality of subjects. (34), and a learning unit (35) that learns a determination algorithm D2 based on the sound information and the diagnostic information of each subject." (see summary).

WO2019/073900

The conventional technology extracts predefined attribute information from the data of the analyzed case and the data of the case to be analyzed, obtains the degree of similarity between the analyzed case and the case to be analyzed, Extract examples. However, since the attribute information used to calculate the degree of similarity is selected by the designer, the attribute information for accurately extracting analyzed cases similar to the analysis target case is not necessarily selected. . Furthermore, there is a high possibility that the attribute information selected by the designer can express only a part of the properties of the case data. In other words, it was difficult to design similarity in the conventional technology.

Further, the technique described in Patent Document 1 learns a determination algorithm using a predetermined type of modeling and feature amounts, but does not have a mechanism for efficient learning using past modeling results. Accordingly, an object of one aspect of the present invention is to accurately and efficiently identify an analyzed case similar to an analysis target case in order to analyze the analysis target case with high accuracy.

In order to solve the above problems, one embodiment of the present invention employs the following configuration. The analysis support device has a processor and a memory, and the memory stores analysis target case data indicating explanatory variables and objective variables of analysis target cases, a model obtained by analyzing the analyzed cases, and parameters applied to the model. and an evaluation value of the model when the analyzed case is analyzed by the model to which the parameter is applied, and the processor retains analysis evaluation data indicating a combination of calculating the evaluation value of the model when predicting the objective variable of the analysis target case by applying a combination of a predetermined model and parameters of the analysis target case to the explanatory variables of the analysis target case, and calculating the calculated evaluation A similarity is calculated by comparing the value and each evaluation value indicated by the analysis evaluation data, and a similar case that is an analyzed case similar to the analysis target case is specified based on the calculated similarity. .

According to one aspect of the present invention, in order to analyze a case to be analyzed with high accuracy, an analyzed case similar to the case to be analyzed can be specified with high accuracy and efficiency.

Problems, configurations, and effects other than those described above will be clarified by the following description of the embodiments.

1 is a block diagram showing a configuration example of an analysis support device according to Embodiment 1; FIG. It is an example of an input table in Example 1. FIG. It is an example of the analysis information in Example 1. FIG. 1 is an example of an analysis database in Example 1. FIG. 4 is an example of a parameter database in Example 1. FIG. 6 is a flow chart showing an example of automatic analysis execution processing in Embodiment 1. FIG. 4 is a flow chart showing an example of rule-based problem classification processing in Example 1. FIG. 7 is a flow chart showing an example of common parameter search processing in Embodiment 1. FIG. 10 is a flowchart illustrating an example of similar case extraction processing according to the first embodiment; FIG. 10 is an explanatory diagram showing a specific example of similar case extraction processing according to the first embodiment; 5 is a flowchart showing an example of parameter recommendation processing in Example 1. FIG. 6 is a flowchart showing an example of peripheral search processing according to the first embodiment; 7 is a flowchart illustrating an example of re-learning processing in Example 1; 5 is a flow chart showing an example of analysis database creation processing in Example 1. FIG. 4 is an example of a display screen displayed on the output device when automatic analysis execution processing is performed in Embodiment 1. FIG.

BEST MODE FOR CARRYING OUT THE INVENTION Hereinafter, embodiments of the present invention will be described in detail based on the drawings. In this embodiment, in principle, the same components are denoted by the same reference numerals, and repeated descriptions are omitted. It should be noted that the present embodiment is merely an example for realizing the present invention and does not limit the technical scope of the present invention.

FIG. 1 is a block diagram showing a configuration example of an analysis support device. The analysis support apparatus 100 includes, for example, a CPU (Control Processing Unit) 101, a memory 102, an auxiliary storage device 103, an input device 104, an output device 105, and a communication device, which are connected to each other by an internal communication line such as a bus 107. 106.

CPU 101 includes a processor and executes programs stored in memory 102 . The memory 102 includes ROM (Read Only Memory), which is a non-volatile storage element, and RAM (Random Access Memory), which is a volatile storage element. The ROM stores immutable programs (for example, BIOS (Basic Input/Output System)) and the like. The RAM is a high-speed and volatile storage device such as a DRAM (Dynamic Random Access Memory), and temporarily stores programs executed by the CPU 101 and data used when the programs are executed.

The auxiliary storage device 103 is, for example, a magnetic storage device (HDD (Hard Disk Drive)), a flash memory (SSD (Solid State Drive)), or other large-capacity and non-volatile storage device. Stores data used when running That is, the program is read from the auxiliary storage device 103, loaded into the memory 102, and executed by the CPU 101. FIG.

The input device 104 is a device that receives input from an operator, such as a keyboard and a mouse. The output device 105 is a device that outputs the execution result of the program in a format that can be visually recognized by the operator, and is, for example, a display or a printer. A communication device 106 is a network interface device that controls communication with other devices according to a predetermined protocol.

Programs executed by the CPU 101 are provided to the analysis support apparatus 100 via removable media (CD-ROM, flash memory, etc.) or a network, and stored in the non-volatile auxiliary storage device 103, which is a non-temporary storage medium. Therefore, the analysis support apparatus 100 preferably has an interface for reading data from removable media.

The analysis support apparatus 100 is a computer system configured on one physical computer or on a plurality of computers configured logically or physically, and operates on the same computer with separate threads. Alternatively, it may operate on a virtual computer built on a plurality of physical computer resources.

The CPU 101 includes, for example, an automatic analysis execution unit 111, an analysis database creation unit 112, a rule-based problem classification unit 113, a common parameter search unit 114, a similar case extraction unit 115, a parameter recommendation unit 116, a peripheral search unit 117, and a relearning unit. 118.

The automatic analysis execution unit 111 controls automatic analysis processing of analysis target cases indicated by an input table 121, which will be described later. The analysis database creation unit 112 creates an analysis database 123, which will be described later. The rule-based problem classification unit 113 identifies a problem to be solved by the model in the analysis target case indicated by the input table 121 .

The common parameter search unit 114 applies a model to which some parameters indicated by the analysis database 123 are applied to the explanatory variables of the input table 121 . The similar case extraction unit 115 extracts similar cases, which are previously analyzed cases similar to the analysis target case.

The parameter recommendation unit 116 recommends a parameter showing the best evaluation value when applied to the analysis target case among the parameters already applied to the similar cases. The surrounding search unit 117 searches for parameters around the recommended parameter, and outputs the best parameter indicating the best evaluation value when applied to the case to be analyzed among the searched parameters. The relearning unit 118 uses the best parameters to generate a model 124, which will be described later, for analyzing the case to be analyzed.

For example, the CPU 101 operates according to an automatic analysis execution program loaded into the memory 102 to function as an automatic analysis execution unit 111, and operates according to an analysis database creation program loaded into the memory 102 to operate as an analysis database creation unit. 112.

Some or all of the functions of the functional units included in the CPU 101 may be realized by hardware such as ASIC (Application Specific Integrated Circuit) or FPGA (Field-Programmable Gate Array).

The auxiliary storage device 103 holds an input table 121, analysis information 122, an analysis database 123, a model 124, and a parameter database 125, for example. Part or all of the information stored in the auxiliary storage device 103 may be stored in the memory 102 or may be stored in a database or the like connected to the analysis support device 100 .

The input table 121 holds data of cases to be analyzed. Analysis information 122 holds additional information for analyzing input table 121 . The analysis database 123 holds information indicating models and parameters used for analysis in cases that have been analyzed in the past, and evaluation values. As will be described later, the analysis database 123 does not need to have information that directly indicates the analyzed case itself (attribute information of the analyzed case).

The model 124 includes models for analyzing cases to be analyzed indicated by the input table 121 and models for analyzing cases that have been analyzed in the past. The parameter database 125 holds information indicating parameters applied to the model 124 .

In this embodiment, the information used by the analysis support apparatus 100 may be represented by any data structure without depending on the data structure. Although the information is represented in the form of a table in this embodiment, the information can be stored in a data structure suitably selected from, for example, a list, database, or queue.
FIG. 2 is an example of the input table 121. As shown in FIG. The input table 121 is table data in which the values of the same variables are stored in the column direction and the values of the respective variables are stored in the row direction.

In the example of FIG. 2, the input table 121 includes a PassengerId column 1211, a Survived column 1212, a Sex column 1213, an Age column 1214, a SibSp column 1215, a Fare column 1216, and an Embarked column 1217. The number of variables (columns) included in the input table 121 is arbitrary, and each variable can take values of various types such as character strings, flags, consecutive numbers (numerical values), and the like.

In the example of FIG. 2, the PassengerId column 1211 is an ID column (that is, a variable for identifying a record), the Survived column 1212 is a target variable column, and the other columns are explanatory variable columns. The purpose of the analysis in this example includes generating a model, which is a prediction formula that more accurately predicts the value of the objective variable column from the value of the explanatory variable column.

In the input table 121 of FIG. 2, the Survived column 1212 is the objective variable column, and the value of the Survived column 1212 takes either 1 or 0, so predicting the value of the Survived column 1212 is a binary classification problem. . Since the ID column is generally a serial number that identifies a data sample, it is often not directly used for analysis.

FIG. 3 is an example of the analysis information 122. As shown in FIG. The analysis information 122 includes, for example, an ID column 1221, a target variable name column 1222, a question column 1223, and a type column 1224. The ID column 1221 holds IDs that identify records of the analysis information 122 . A target variable name column 1222 indicates the target variable column of the input table 121 . A question column 1223 indicates which class of problem the input table 121 is classified into, such as binary classification, multiclass classification, or regression. A type column 1224 indicates the type of the objective variable column.

FIG. 4 is an example of the analysis database 123. As shown in FIG. The analysis database 123 is defined for each problem classification, and FIG. 4 shows an example of the analysis database 123 whose problem classification is multi-class problem and rare phenomenon problem. The analysis database 123 includes, for example, an Analysis_ID column 1231, a Model_Name column 1232, a Common_Search column 1233, a Recipe_ID column 1234, a Best_Flag column 1235, and an Accuracy column 1236, and holds past search result information.

The Analysis_ID column 1231 holds IDs that identify analysis cases (that is, IDs that are assigned when one input table 121 is analyzed). Multiple results can be stored for one Analysis_ID because one or more searches and evaluations are performed with different parameters for one analysis case.

A Model_Name column 1232 holds the name of the model 124 used in the analysis. Return to Logistic, Return to Logistic, XG Boost, RANDOM FOREST, SVM (SUPPORT VECTOR MACHINE), SVR (SUPPPORT VECTOR REGRESSION), NEURAL NETWORK, And Gaussian Process is an example of a model 124 used for analysis.

The Common_Search column 1233 holds flags for identifying parameters used to extract analyzed cases similar to the analysis target cases indicated by the input table 121 .

The analysis support apparatus 100 of the present embodiment extracts analyzed cases similar to an analysis target case, and recommends parameters that enable highly accurate analysis when used to analyze the analysis target case. In order to find similar cases, the analysis support apparatus 100 can compare cases by making the search target parameters the same. Therefore, the analysis database 123 has a Common_Search column 1233 as a flag for identifying samples searching for the same parameter among analysis cases.

The Recipe_ID column 1234 holds Recipe_ID, which is an ID for identifying each searched parameter set. When a parameter set included in the analysis database 123 is recommended, the parameter set is designated by this Recipe_ID.

Best_Flag column 1235 holds Best_Flag, which is a flag for identifying the parameter set for which the model performed best in the same analysis case, ie, the case with the same Analysis_ID. In the example of FIG. 4, the performance of the model is evaluated by the prediction accuracy (the value of the Accuracy column 1236), but it is possible to evaluate the performance of the model from various viewpoints such as other evaluation indexes and model stability. .

The Accuracy column 1236 indicates the prediction accuracy (Accuracy), which is an example of the evaluation value of the model when the parameter set indicated by the Recipe_ID column 1234 is applied to the model. Although the analysis database 123 is named database for convenience of explanation, it does not necessarily have to be a database, and may be in a text format or other format, for example.

FIG. 5 is an example of the parameter database 125. As shown in FIG. The parameter database 125 includes, for example, a Recipe_ID column 1251 that holds Recipe_IDs. The parameter database 125 also includes, for example, a param_max_depth column 1252, a param_leafe_num column 1253, a param_n_estimator column 1254, and a param_learning_rate column 1255, all of which hold parameter values. In the example of FIG. 5, the parameter database 125 holds four parameters, but the types and number of parameters held by the parameter database 125 depend on the model 124 that the analysis support apparatus 100 has.

It should be noted that the illustrated columns of each data in the auxiliary storage device 103 are merely examples, and each table does not necessarily have all the illustrated columns, and may have additional columns. may

FIG. 6 is a flowchart showing an example of automatic analysis execution processing. It is assumed that the values of the input table 121 and the analysis information 122 have been input before the processing of FIG. However, the problem column 1223 of the analysis information 122 does not have to store a value (a null value may be stored).

In the automatic analysis execution process, the automatic analysis execution unit 111 searches the analysis database 123 for past analyzed cases similar to the analysis target case based on the input table 121 , the analysis information 122 and the parameter database 125 . Then, the automatic analysis execution unit 111 extracts a parameter indicating the best evaluation value, applies the extracted parameter, and outputs a model for analyzing the analysis target case.

Since the details of each step in FIG. 6 will be described later, they will be briefly described here. First, the rule-based problem classification unit 113 determines to which problem the problem solved by the model in the analysis target case indicated by the input table 121 belongs (S601). The common parameter search unit 114 calculates the evaluation value of the model when the model using the parameters applied to past analyzed cases is applied to the explanatory variables of the input table 121 (S602).

The similar case extracting unit 115 compares the evaluation value in the input table 121, which is the analysis target case to which the parameter is applied, with the evaluation values of the past analyzed cases, and performs the analysis most similar to the analysis target case. Similar cases, which are completed cases, are extracted (S603).

The parameter recommendation unit 116 extracts and recommends the parameter with the highest evaluation value among the parameters applied to the similar cases (S604). The surrounding search unit 117 searches around the recommended parameters and extracts the best parameters (S605). The relearning unit 118 uses the best parameters to generate a model 124 for analyzing the case to be analyzed.

FIG. 7 is a flowchart showing an example of rule-based problem classification processing in step S601. The rule-based problem classification unit 113 reads the input table 121 and analysis information 122 (S701).

The rule-based problem classification unit 113 determines whether a problem classification is defined in the analysis information 122, that is, whether a value is entered in the problem field 1223 (S702).

When the rule-based problem classification unit 113 determines that the problem classification is defined in the analysis information 122 (S702: Yes), it outputs the problem classification indicated by the problem column 1223 as a problem classification result (S707). End the classification process.

When the rule-based problem classification unit 113 determines that the problem classification is not defined in the analysis information 122 (S702: No), it identifies the objective variable column of the input table 121 from the objective variable name column 1222 of the analysis information 122, It is determined whether the number of unique elements (that is, the number of element types) in the objective variable column of the input table 121 is 2 or less (S703).

When the rule-based problem classification unit 113 determines that the number of unique elements in the objective variable column is 2 or less (S703: Yes), the problem classification result , executes the rare phenomenon determination process (S705), outputs the problem classification result (S707), and terminates the rule-based problem classification process.

Note that, in step S705, the rule-based problem classification unit 113 includes information indicating that the problem in the input table 121 includes a rare phenomenon problem in the problem classification result when the bias in the values of the objective variable column is large. Specifically, for example, the rule-based problem classification unit 113 determines whether the variance of the objective variable column values is equal to or greater than a predetermined value, or when the difference between the maximum relative frequency and the minimum relative frequency of the objective variable column values is If the value is equal to or greater than a predetermined value, etc., it is determined that the bias in the values of the objective variable column is large.

When the rule-based problem classification unit 113 determines that the number of unique elements in the objective variable column exceeds 2 (S703: No), it determines whether the objective variable is character data (S705). Specifically, for example, the rule-based problem classification unit 113 may refer to the value of the type column 1224 of the analysis information 122 to determine whether or not the objective variable is character data. If at least one character data is included in the value of the objective variable column, it may be determined that the objective variable is character data. Thus, the rule-based problem classification unit 113 executes problem classification based on the data type of the objective variable.

When the rule-based problem classification unit 113 determines that the objective variable is character data (S704: Yes), the problem classification result includes information indicating that the problem in the input table 121 includes a multi-class problem, and the rare phenomenon is determined. The process is executed (S705), the problem classification result is output (S707), and the rule-based problem classification process ends.

When the rule-based problem classification unit 113 determines that the objective variable is not character data (that is, is numerical data) (S704: No), the number of unique elements in the objective variable column of the input table 121 is equal to or less than the predetermined threshold α. (S705).

When the rule-based problem classification unit 113 determines that the number of unique elements in the objective variable column is equal to or less than α (S705: Yes), the problem classification result includes information indicating that the problem in the input table 121 includes a multi-class problem. The rare phenomenon determination process is executed (S705), the problem classification result is output (S707), and the rule-based problem classification process ends. When the rule-based problem classification unit 113 determines that the number of unique elements in the objective variable column exceeds α (S705: No), the problem classification result includes information indicating that the problem in the input table 121 is a regression problem, The problem classification result is output (S707), and the rule-based problem classification process ends.

By the processing of FIG. 7, the problem classification results of the input table 121 include the problem indicated by the problem column 1223 of the analysis information 122, the binary problem, the binary problem and rare phenomenon problem, the multiclass problem, the multiclass problem and rare phenomenon problem, or regression problem.

FIG. 8 is a flowchart showing an example of common parameter search processing in step S602. The common parameter searching unit 114 reads out the input table 121, the analysis information 122, and the analysis database 123 corresponding to the problem indicated by the problem classification result output in step S707 (S801).

The common parameter search unit 114 identifies the model and Recipe_ID of the record whose value in the Common_Search column 1233 of the analysis database 123 read in step S801 is 1, and extracts the parameter set corresponding to the identified Recipe_ID from the parameter database 125 as common_grid. Extract (S802).

For each parameter set extracted as common_grid, the common parameter search unit 114 applies a model corresponding to the parameter set in the analysis database 123 to the values of the explanatory variables in the input table 121, and the model to which the parameter set is applied is An evaluation value (for example, prediction accuracy) of the model when applied to the explanatory variables of the input table 121 is calculated (S803).

The common parameter search unit 114 outputs the searched parameter group θ _search , which is the parameter set extracted in step S802, in association with the evaluation value obtained in step S803 (S804).

FIG. 9 is a flowchart showing an example of similar case extraction processing in step S603. The similar case extraction unit 115 extracts the input table 121, the analysis information 122, the searched parameter group θ _search output in step S804, and the analysis database 123 corresponding to the problem indicated by the problem classification result output in step S707. , is read out (S901).

The similar case extraction unit 115 compares the evaluation value (Accuracy) for common_grid with the evaluation value in the searched parameter group θ _search for each Analysis_ID of the analyzed case in the analysis database 123, and compares the analyzed case with The degree of similarity with the analysis target case in the input table 121 is calculated (S902).

In step S902, the similar case extraction unit 115 _calculates , for example, the correlation coefficient, Euclidean distance, Manhattan distance, or absolute error or the like is calculated as the degree of similarity. The similar case extraction unit 115 outputs the Analysis_ID of the analyzed case with the highest similarity to the analysis target case in the input table 121 as the similar case ID (S903).

FIG. 10 is an explanatory diagram showing a specific example of the similar case extraction processing shown in FIG. As described above, the similar case extraction unit 115 extracts Accuracy, which is the evaluation value of the model when the searched parameter group θ _search is applied, in the analysis target cases obtained from the input table 121, and the analyzed cases in the analysis database 123. The degree of similarity between Accuracy, which is the evaluation value of the parameter set corresponding to θ _search in each case, is calculated. As described above, the analysis database 123 stores a plurality of analyzed cases and is given an Analysis_ID for identifying the analyzed cases.

Correlation coefficients between the evaluation values of the cases with Analysis_ID of "1", "2", "3", and "4" and the evaluation values of the analysis target cases obtained from the input table 121 are r= 0.11, r=0.32, r=0.25, r=0.85. Therefore, in the example of FIG. 10, it can be seen that the analyzed case with the highest degree of similarity to the analysis target case is the analyzed case with Analysis_ID of "4" showing r=0.85.

Therefore, in the example of FIG. 10, the similar case extraction unit 115 outputs "4" as the similar case ID. When calculating the correlation coefficient as the degree of similarity, the similar case extraction unit 115 may output the Analysis_ID of the analyzed case with the largest absolute value of the correlation coefficient as the similar case ID.

Through the above-described processing, the analysis support apparatus 100 calculates the evaluation value of the model in the analysis target case indicated by the input table 121 only for common_grid among the analysis results of the analyzed cases, and calculates the calculated evaluation value and the past analyzed case. Since similar cases are extracted by comparing the evaluation value of the model in and , similar cases can be extracted with a small amount of processing.

In addition, in the above-described processing, the analysis support apparatus 100 uses the analysis database 123 (the models and parameters of the analyzed cases, and the evaluation values of the models) without using the data of the analyzed cases themselves (the attribute information of the analyzed cases). ) can be used to extract similar cases, the amount of data can be reduced, and even if the analysis database 123 is leaked, personal information and the like will not be leaked.

FIG. 11 is a flowchart showing an example of parameter recommendation processing in step S604. The parameter recommendation unit 116 reads out the input table 121, the analysis information 122, the similar case ID output in step S904, the parameter database 125, and the analysis database 123 corresponding to the problem classification results (S1101).

The parameter recommendation unit 116 refers to the analysis database 123 corresponding to the problem classification result, and searches for the highest Accuracy evaluation value of the analyzed case indicated by the similar case ID as the best evaluation value (S1102).

The parameter recommendation unit 116 acquires the Recipe_ID of the record indicating the best evaluation value, and acquires the parameter set corresponding to the Recipe_ID from the parameter database 125 (S1103). The parameter recommendation unit 116 outputs the acquired parameter set as a recommended parameter θ _recommend (S1104), and terminates the parameter recommendation process.

In the example of FIG. 11, the parameter recommendation unit 116 recommends one parameter set showing the highest evaluation value in similar cases. It may be output as _θrecommend , or a parameter set selected from a viewpoint other than the numerical value of the evaluation value may be output as _{the recommended} parameter θrecommend.

Further, the parameter recommendation unit 116 may, for example, output a predetermined number (for example, 10) of parameter sets in descending order of evaluation values in similar cases as recommended parameters _θrecommend , or output predetermined parameters in descending order of evaluation values in similar cases. A number (for example, 10) of parameter sets may be output as _{the recommended} parameter θrecommend.

Further, the parameter recommendation unit 116 may output, for example, all parameter sets with evaluation values equal to or greater than a predetermined value in similar cases as recommended parameters θ _recommend . All parameter sets that are less than a predetermined value may be output as _recommended parameters θrecommend.

FIG. 12 is a flow chart showing an example of the peripheral search processing in step S605. The surrounding search unit 117 reads the input table 121, the analysis information 122, and _{the recommendation} parameter θrecommend output in step S1104 (S1201).

The peripheral search unit 117 uses the recommended parameter _θrecommend as an initial value, applies a model to which the parameter set around the _recommended parameter θrecommend is applied to the explanatory variables of the input table 121, and searches for the parameter set showing the best evaluation value. (S1202). The surrounding search unit 117 outputs the parameter set indicating the best evaluation value as the best parameter θ _best (S1203), and ends the search process.

In this way, the peripheral searching unit 117 can acquire the best parameter θ _best indicating a highly accurate result with a small amount of processing by searching around the recommended parameter θ _recommend in similar cases.

In step S1202, peripheral searching section 117 searches for a parameter set within a predetermined range, for example.

Further, in step S1202, the surrounding search unit 117 may search for a parameter set in an updated (for example, enlarged or reduced) range based on a predetermined condition centering on the recommended parameter _θrecommend , for example.

Specifically, for example, the peripheral searching unit 117 calculates var=(x _max −x _min )×(1−|Corr|) (where x _max and x _min are each within a predetermined range of parameters). are the maximum value and the minimum value, and Corr is the correlation coefficient between the similar case and the analysis target case indicated by the input table 121). Peripheral search unit 117 sets x _{new_min} = x _recommend −var as the minimum value of the updated parameter range and x _{new_max} = x _recommend + var as the maximum value of the updated parameter range, and searches for a parameter set in the range. (However, x _recommend is a recommendation parameter).

FIG. 13 is a flow chart showing an example of the relearning process in step S606. The relearning unit 118 reads the input table 121, the analysis information 122, and the best parameter θ _best output in step S1203 (S1301).

The relearning unit 118 performs preprocessing on the input table 121 based on the best parameter θ _best (S1302). Specifically, for example, the relearning unit 118 performs normalization of numeric columns, integration of notational variations in character string columns, dimension compression using PCA (Principal Component Analysis), etc., conversion of character string columns into dummy variables, Both outlier processing and outlier processing are examples of preprocessing.

The relearning unit 118 generates the feature amount of the input table 121 according to the best parameter θ _best (S1303). Specifically, for example, the re-learning unit 118 generates feature values by dividing numerical columns, logarithmic conversion, exponential conversion, time-series feature value conversion, and/or conversion of time data into date.

The relearning unit 118 performs modeling on the input table 121 using the best parameter θ _best (S1304). Linear Regression, Logistic Regression, SVM, SVR, GaussianProcess, RandomForest, LightGBM, XGBoost, and NeuralNetwork are all examples of modeling. The relearning unit 118 outputs the model 124 for predicting the target variable of the example in the input table 121 obtained in the modeling in step S1304 and the analysis result of the model 124 (S1305), and ends the relearning process.

FIG. 14 is a flowchart showing an example of analysis database creation processing. Since the analysis database creation process is a process of accumulating the analysis results for the input table 121 in the analysis database 123, it is desirable to be performed after the automatic analysis execution process of FIG. 6 is completed.
The analysis database creating unit 112 reads the input table 121 and the analysis information 122 (S1401).

The analysis database creation unit 112 executes common parameter search processing (S1402). In step S1402, the analysis database creation unit 112 searches the input table 121 for a parameter set in the same manner as in the common parameter search process of FIG.

The analysis database creation unit 112 executes detailed search (S1403). In step S1403, the analysis database creation unit 112 searches for parameter sets (for example, all parameter sets within a predetermined range from each common_grid) other than the common parameters (that is, common_grid) between the analysis cases searched in step S1402.

The analysis database creation unit 112 assigns 1 as Best_Flag to the sample showing the highest evaluation value among all the parameter sets searched in steps S1302 and S1303 (S1404). Note that the process of step S1404 is a process for speeding up the operation during operation, so it does not necessarily have to be executed.

The analysis database creation unit 112 assigns Analysis_ID to the cases in the input table 121, records Analysis_ID, Model_Name, Recipe_ID, Best_Flag, and Accuracy in the analysis database 123 (S1404), and ends the analysis database creation process.

When searching for a parameter set that is not stored in the parameter database 125, the analysis database creation unit 112 assigns a Recipe_ID to the parameter set, and records the Recipe_ID and the parameter set in the parameter database 125. .

In addition, the value of the Common_Search column 1233 of the analysis database 123 may be recorded, for example, according to the input of the user of the analysis support apparatus 100, or automatically (for example, among predetermined models, for records indicating Accuracy equal to or greater than a predetermined value 1 for other records, 0) may be recorded.

FIG. 15 is an example of a display screen displayed on the output device 105 when automatic analysis execution processing is performed. The display screen 1500 includes, for example, an output information display area 1501 , a similar case extraction recommendation execution button 1502 , a re-search execution button 1503 , and a model learning execution button 1504 .

When the similar case extraction recommendation execution button 1502 is selected, the automatic analysis execution processing of FIG. The degree of similarity between the target case and the similar case and the recommended parameters output in the parameter recommendation process of FIG. 11 are displayed.

When the re-search execution button 1503 is selected, the surrounding search processing of FIG. 12 is executed and the best parameters are further displayed in the output information display area 1501 . When the model learning execution button 1504 is selected, the relearning process of FIG. 13 is executed and the output model is further displayed.

In addition, the present invention is not limited to the above-described embodiments, and includes various modifications. For example, the above-described embodiments have been described in detail in order to explain the present invention in an easy-to-understand manner, and are not necessarily limited to those having all the described configurations. It is also possible to replace part of the configuration of one embodiment with the configuration of another embodiment, or to add the configuration of another embodiment to the configuration of one embodiment. Moreover, it is possible to add, delete, or replace a part of the configuration of each embodiment with another configuration.

Further, each of the above configurations, functions, processing units, processing means, and the like may be realized by hardware, for example, by designing them in an integrated circuit. Moreover, each of the above configurations, functions, etc. may be realized by software by a processor interpreting and executing a program for realizing each function. Information such as programs, tables, and files that implement each function can be stored in recording devices such as memories, hard disks, SSDs (Solid State Drives), or recording media such as IC cards, SD cards, and DVDs.

Further, the control lines and information lines indicate those considered necessary for explanation, and not all control lines and information lines are necessarily indicated on the product. In practice, it may be considered that almost all configurations are interconnected.

100 analysis support device, 101 CPU, 102 memory, 103 auxiliary storage device, 104 input device, 105 output device, 106 communication device, 111 automatic analysis execution unit, 112 analysis database creation unit, 113 rule-based problem classification unit, 114 common parameter search unit 115 similar case extraction unit 116 parameter recommendation unit 117 peripheral search unit 118 relearning unit 121 input table 122 analysis information 123 analysis database 124 model 125 parameter database

Claims (9)

An analysis support device,
having a processor and a memory,
The memory is
analysis target case data indicating explanatory variables and objective variables of analysis target cases;
analysis evaluation data indicating a combination of a model that analyzed an analyzed case and parameters applied to the model, and an evaluation value of the model when the analyzed case was analyzed by the model to which the parameters were applied; , and hold
The processor
Calculating the evaluation value of the model when predicting the objective variable of the analysis target case by applying a combination of predetermined models and parameters included in the analysis evaluation data to the explanatory variables of the analysis target case death,
calculating a degree of similarity by comparing the calculated evaluation value with each evaluation value indicated by the analysis evaluation data;
An analysis support device that identifies a similar case, which is an analyzed case similar to the analysis target case, based on the calculated degree of similarity. The analysis support device according to claim 1,
The processor refers to the analysis evaluation data, and determines, as a recommended parameter, the parameter in the combination with the highest evaluation value among the combinations of the model that analyzed the similar case and the parameter applied to the model. Device. The analysis support device according to claim 2,
having a display device;
The analysis support device, wherein the processor displays information indicating the similar cases, the degree of similarity, and the recommended parameters on the display device. The analysis support device according to claim 2,
The processor
searching for parameters in a search range that includes the recommended parameters and satisfies a predetermined condition;
Among the searched parameters, the parameter with the highest evaluation value of the model when predicting the target variable of the case to be analyzed by the model corresponding to the recommended parameter to which the searched parameter is applied is determined as the best parameter. analysis support device. The analysis support device according to claim 4,
The processor
calculating a correlation coefficient between the calculated evaluation value and the evaluation value of the similar case;
An analysis support device that determines the search range so that the search range becomes smaller as the absolute value of the correlation coefficient increases. The analysis support device according to claim 4,
The analysis support device, wherein the processor outputs an analysis result obtained by applying the model corresponding to the recommended parameter to which the best parameter is applied to the case to be analyzed. The analysis support device according to claim 1,
The analysis support device, wherein the processor stores, in the analysis evaluation data, a combination of the partial model that analyzed the analysis target case and the parameter applied to the model, and the calculated evaluation value. The analysis device according to claim 1,
the analytical evaluation data indicates a problem solved by the model in the analyzed cases;
The processor
identifying a problem to be solved by a model in the case to be analyzed based on the number of elements of the objective variable, the data type of the objective variable, and the relative frequency of the value of the objective variable in the case data to be analyzed;
The model when the target variable of the case to be analyzed is predicted by the combination of the model and parameters corresponding to the identified problem, among the combinations of the predetermined partial models and parameters included in the analysis evaluation data An analysis support device that calculates the evaluation value of An analysis support method using an analysis support device,
The analysis support device has a processor and a memory,
The memory is
analysis target case data indicating explanatory variables and objective variables of analysis target cases;
analysis evaluation data indicating a combination of a model that analyzed an analyzed case and parameters applied to the model, and an evaluation value of the model when the analyzed case was analyzed by the model to which the parameters were applied; , and hold
The analysis support method includes:
of the model when the processor predicts the objective variable of the analysis target case by applying a combination of predetermined models and parameters included in the analysis evaluation data to the explanatory variables of the analysis target case; Calculate the evaluation value,
The processor compares the calculated evaluation value with each evaluation value indicated by the analysis evaluation data to calculate a degree of similarity;
An analysis support method, wherein the processor identifies a similar case, which is an analyzed case similar to the case to be analyzed, based on the calculated degree of similarity.

Images