Dataset augmentation assistance device, dataset augmentation assistance method and dataset augmentation assistance program

Abstract

To facilitate proper supplementation of data to a dataset.SOLUTION: There is provided a method for calculating information relating to supplemental data for supplementing an actual measurement dataset including a plurality of items of actual measurement data each composed of a set of values of one or more response variables and values of one or more explanatory variables. A processor executes: clustering processing for clustering a target dataset into a plurality of clusters; supplementation target cluster selection processing for selecting, from the plurality of clusters, at least one supplementation target cluster to be supplemented with the supplemental data; explanatory variable calculation processing for calculating a value of an explanatory variable of the supplemental data on the basis of the selected at least one supplementation target cluster and an actual measurement data model trained using the target dataset and representing relations between the values of the one or more response variables and the values of the one or more explanatory variables of the actual measurement data; and explanatory variable output processing for outputting the calculated value of the explanatory variable of the supplemental data.SELECTED DRAWING: Figure 4

Description

[Technical field]

[0001] The present invention relates to a dataset expansion support device, a dataset expansion support method, and a dataset expansion support program that calculate information about supplementary data to be supplemented to an actual measurement dataset including a plurality of actual measurement data each consisting of a pair of values ​​of one or more explanatory variables and values ​​of one or more objective variables. [Background technology]

[0002] There is a technology that uses a machine learning model or a statistical model to learn actual measurement data consisting of a set of one or more explanatory variable values ​​and one or more objective variable values, and utilizes the learned model. For example, Patent Document 1 describes an analysis system that performs machine learning using an optical signal emitted from an observed object as learning data, and changes a discrimination algorithm that determines whether the observed object is a specific object based on the results of the machine learning. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] JP 2023-98911 A Summary of the Invention [Problem to be solved by the invention]

[0004] Generally, to perform machine learning appropriately, the amount and quality of the learning data used in the machine learning must be sufficient. With the technology described in Patent Document 1, it is not known when the machine learning is sufficient. With the technology described in Patent Document 1, there is a risk of overlearning, which may result in inappropriate machine learning.

[0005] From each piece of actual measurement data, it is not possible to determine whether the actual measurement data is a singular point such as a peak in the data space of the actual measurement data. Furthermore, machine learning models and models of statistical methods learn the properties of the actual measurement data as a group contained in a training dataset used as training data. Therefore, unless the properties of the training dataset used as training data as a set of actual measurement data in which a plurality of pieces of actual measurement data are compiled are taken into consideration, it is not easy to determine whether the training dataset appropriately contains actual measurement data. Furthermore, since it is not easy to take into consideration the properties of the actual measurement data as a set, there is a problem in that it is not easy to appropriately supplement the dataset used as training data with data.

[0006] Therefore, an object of the present invention is to provide a dataset expansion support device, a dataset expansion support method, and a dataset expansion support program that make it easy to appropriately supplement a dataset used for model training with data. [Means for solving the problem]

[0007] In order to achieve the above-mentioned object, one aspect of the dataset expansion support method of the present invention is a dataset expansion support method in a dataset expansion support device which includes a processor and a storage device for storing information used in processing by the processor, and in which the processor calculates information regarding supplementary data to be supplemented to an actual measurement dataset including a plurality of actual measurement data each including a pair of values ​​of one or more objective variables and values ​​of one or more explanatory variables, in which the processor executes a clustering process for clustering a target dataset into a plurality of clusters, a supplementary target cluster selection process for selecting at least one supplementary target cluster to be supplemented with supplementary data from the plurality of clusters, an explanatory variable calculation process for calculating values ​​of explanatory variables of the supplementary data based on an actual measurement data model which is learned using the target dataset and which represents a relationship between values ​​of the one or more objective variables and values ​​of the one or more explanatory variables of the actual measurement data, and the selected at least one supplementary target cluster, and an explanatory variable output process for outputting the calculated values ​​of explanatory variables of the supplementary data.

[0008] Moreover, one aspect of the dataset expansion support device of the present invention is a dataset expansion support device that includes a processor and a storage device that stores information used in the processing of the processor, and the processor calculates information regarding supplementary data to be supplemented to an actual measurement dataset including a plurality of actual measurement data each including a pair of values ​​of one or more objective variables and values ​​of one or more explanatory variables, and the processor executes a clustering process that clusters the target dataset into a plurality of clusters, a supplementary target cluster selection process that selects at least one supplementary target cluster to be supplemented with the supplementary data from the plurality of clusters, an explanatory variable calculation process that calculates values ​​of explanatory variables of the supplementary data based on an actual measurement data model that is learned using the target dataset and that represents a relationship between the values ​​of the one or more objective variables and the values ​​of the one or more explanatory variables of the actual measurement data and the selected at least one supplementary target cluster, and an explanatory variable output process that outputs the calculated values ​​of explanatory variables of the supplementary data.

[0009] Moreover, one embodiment of the dataset expansion support program of the present invention is a dataset expansion support program in a dataset expansion support device that includes a processor and a storage device that stores information used in the processing of the processor, and in which the processor calculates information regarding supplementary data to be supplemented to an actual measurement dataset including a plurality of actual measurement data each consisting of a pair of values ​​of one or more objective variables and values ​​of one or more explanatory variables, and causes the processor to execute a clustering process that clusters a target dataset into a plurality of clusters, a supplementary target cluster selection process that selects at least one supplementary target cluster to be supplemented with supplementary data from the plurality of clusters, an explanatory variable calculation process that calculates values ​​of explanatory variables of the supplementary data based on an actual measurement data model that is learned using the target dataset and that represents the relationship between the values ​​of the one or more objective variables and the values ​​of the one or more explanatory variables of the actual measurement data and the selected at least one supplementary target cluster, and an explanatory variable output process that outputs the calculated values ​​of explanatory variables of the supplementary data. Effect of the Invention

[0010] Exemplary aspects of the present invention facilitate appropriate supplementation of a data set used to train a model.

[0011] Other objects, configurations and advantages than those mentioned above will become apparent from the following description of the embodiments. [Brief description of the drawings]

[0012] [Figure 1] FIG. 1 is a diagram illustrating an example of a functional block diagram of a dataset expansion support device according to an embodiment. [Diagram 2] FIG. 2 is a block diagram illustrating an example of a hardware configuration of a dataset expansion support device according to an embodiment. [Diagram 3] FIG. 3 is an explanatory diagram showing an example of a condition input screen displayed on a user terminal. [Figure 4] FIG. 4 is a flowchart illustrating an example of a data set expansion support process according to the embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0013] Hereinafter, an embodiment of the present invention will be described with reference to the drawings. However, the present invention is not to be interpreted as being limited to the description of the embodiment shown below. It will be easily understood by those skilled in the art that the specific configuration can be changed without departing from the concept or purpose of the present invention.

[0014] In the configurations of the invention described below, the same or similar configurations or functions are given the same reference numerals, and duplicated explanations are omitted.

[0015] In this specification, the terms "first," "second," "third," and the like are used to identify components and do not necessarily limit the number or order.

[0016] In this specification, as an example of various information, the term "XX table" may be used, but it may also be expressed as a data structure such as an "XX list" or "XX queue." Furthermore, the "XX table" may also be expressed as "XX information." When explaining identification information, the terms "identification information," "identifier," "name," "ID," "number," and the like are used, but these are interchangeable. EXAMPLES

[0017] <<System configuration>> Fig. 1 is a diagram showing an example of a functional block diagram of a dataset expansion support device 1 in an embodiment. As shown in Fig. 1, the dataset expansion support device 1 is connected to a user terminal 500 via a network NW.

[0018] The user terminal 500 stores an actual measurement dataset including a plurality of actual measurement data sets each consisting of a pair of one or more explanatory variable values ​​and one or more objective variable values. The user terminal 500 can transmit and receive information to and from the dataset expansion support device 1 via the network NW.

[0019] The user terminal 500 is a measuring device control device that is attached to a measuring device and controls the measuring device to make the measuring device perform measurements. The user terminal 500 may also be a personal computer or the like. The actual measurement data is, for example, experimental data measured by a measuring device. The measuring device is, for example, a measuring device for chromatography, capillary electrophoresis, photoelectron spectroscopy, infrared absorption spectroscopy, nuclear magnetic resonance spectroscopy, fluorescence spectroscopy, X-ray fluorescence spectroscopy, visible / ultraviolet absorption spectroscopy, Raman spectroscopy, atomic absorption spectroscopy, flame emission spectroscopy, emission spectroscopy, X-ray absorption spectroscopy, X-ray diffraction, electron spin resonance spectroscopy, mass spectrometry, or the like. These measuring devices can obtain actual measurement data in which parameters related to the sample, measurement conditions, etc. are the values ​​of one or more explanatory variables, and the numerical values ​​of the measurement results are the values ​​of one or more objective variables.

[0020] The actual measurement data may be data consisting of a set of one or more explanatory variable values ​​and one or more objective variable values. The actual measurement data may be actual measurement data in which at least one explanatory variable is a parameter of a physical and chemical action on a sample that is a product of a creation process or a manufacturing process (e.g., a manufacturing process of a semiconductor manufacturing device using a plasma process) in which a sample is generated by applying physical and chemical actions such as temperature (e.g., high temperature), pressure, atmospheric pressure (e.g., vacuum), and light irradiation to raw materials, materials, parts, or assemblies thereof, and at least one objective variable is data of physical properties.

[0021] The network NW may be a wired network or a wireless network. The network NW may be a global network such as the Internet or a local area network (LAN).

[0022] The dataset expansion support device 1 includes a storage unit 11 and a processing unit 12. The storage unit 11 stores data used by the processing unit 12. The processing unit 12 includes a data response unit 121 and a dataset expansion support unit 122.

[0023] The data response unit 121 receives the actually measured dataset and the like from the user terminal 500 (arrow A in FIG. 1). The dataset expansion support unit 122 performs a dataset expansion process for evaluating the actually measured dataset received from the user terminal 500, and outputs evaluation result information for the actually measured dataset, such as information about data that is desirable to supplement the actually measured dataset, to the data response unit 121. The data response unit 121 transmits the evaluation result information for the actually measured dataset acquired from the dataset expansion support unit 122 to the user terminal 500 (arrow B in FIG. 1).

[0024] Fig. 2 is a block diagram showing an example of a hardware configuration of the dataset expansion support device 1. As shown in Fig. 2, the dataset expansion support device 1 has a processor 21, a main storage device 22, a sub-storage device 23, an input device 24, an output device 25, a network I / F 26, and a bus 27 connecting these devices.

[0025] The dataset expansion support device 1 can be realized by a general information processing device such as a PC or a server computer. The user terminal 500 can also be realized by using the same hardware resources as the dataset expansion support device 1.

[0026] The processor 21 reads data and programs stored in the secondary storage device 23 into the main storage device 22, and executes the processing defined by the programs.

[0027] The main memory device 22 includes a volatile memory element such as a RAM, and stores the programs executed by the processor 21 and data.

[0028] The secondary storage device 23 is a device having a non-volatile storage element, such as a hard disk drive (HDD) or a solid state drive (SSD), and stores programs, data, etc. The secondary storage device 23 stores an actual measurement model store 231. The actual measurement model store 231 stores various types of models that represent the relationship between the values ​​of one or more explanatory variables of the actual measurement data and the values ​​of one or more objective variables of the actual measurement data.

[0029] The actual measured data model (including the pre-learning actual measured data model) may be a model that represents the relationship between the values ​​of one or more explanatory variables of the actual measured data and the values ​​of one or more objective variables of the actual measured data. The actual measured data model may be a machine learning model such as a support vector machine, a neural network, a decision tree, or clustering. The actual measured data model may also be a model of multivariate analysis of statistical analysis such as regression analysis, correlation analysis, principal component analysis, or multiple regression analysis. Furthermore, the actual measured data model may be a program that represents the relationship between the values ​​of one or more explanatory variables of the actual measured data and the values ​​of one or more objective variables of the actual measured data.

[0030] In addition, a data response program 121a and a dataset expansion support program 122a are installed in the sub-storage device 23. The data response unit 121 and the dataset expansion support unit 122 described above with reference to Fig. 1 are realized by the processor 21 reading the data response program 121a and the dataset expansion support program 122a stored in the sub-storage device 23 into the main storage device 22 and executing them. The processor 21 is the processing unit 12.

[0031] The input device 24 is a device such as a keyboard or a mouse that accepts user operations and acquires information input by the user's operations. The output device 25 is a device such as a display that outputs information and presents information to the user by displaying it on a screen, for example.

[0032] The network I / F 26 is an interface for transmitting and receiving data to and from devices such as the user terminal 500 via the network NW. In other words, the network I / F 26 is a transmitting / receiving device capable of transmitting and receiving information to and from the user terminal 500 via the network NW. The dataset expansion support device 1 can use the network I / F 26 to transmit and receive data to and from devices such as the user terminal 500 connected to the network NW.

[0033] <<Processing Procedure>> Next, a processing procedure of the dataset expansion support device 1 will be described. A user operates the user terminal 500 to cause the user terminal 500 to access the dataset expansion support device 1. When the dataset expansion support device 1 is accessed from the user terminal 500, the data response unit 121 transmits condition input screen information to the user terminal 500. The condition input screen information includes information on the configuration of the condition input screen and information to the effect that the condition input screen is to be displayed on the user terminal 500. The condition input screen is a screen for inputting information used in the dataset expansion process performed by the dataset expansion support unit 122.

[0034] Fig. 3 is an explanatory diagram showing an example of a condition input screen 300 displayed on the user terminal 500. The condition input screen 300 shown in Fig. 3 has an input dataset storage location input field 301, a clustering method selection button 302, a model selection button 303, an explanatory variable setting value input table 310, a row deletion button 315, a row addition button 316, an objective variable setting value input table 320, a row deletion button 325, a row addition button 326, and a send button 330.

[0035] The input dataset storage location input field 301 is a field for inputting the location (folder, directory) where the dataset to be analyzed (input dataset) is stored.

[0036] The clustering method selection button 302 is a button for selecting a method of clustering a target dataset including a plurality of actual measurement data to be analyzed in the dataset expansion support process described later with reference to Fig. 4. When the clustering method selection button 302 is pressed, a list of clustering methods is displayed so that the clustering method can be selected. In the dataset expansion support process described later with reference to Fig. 4, (A) clustering based on data density and (B) clustering based on the value of a response variable are shown as examples of clustering methods.

[0037] (A) Clustering by data density is used when supplementing data to a target dataset for model training so that a model can be created that accurately represents the relationship between the values ​​of one or more explanatory variables of the measured data and the values ​​of one or more objective variables throughout the entire data space of the measured data. The data space of the measured data is a space in which at least one explanatory variable and at least one objective variable of the measured data are taken as coordinates. In the data space of the measured data, parts where data density is low can be considered to be lacking in data. In (A) Clustering by data density, clustering is performed by data density, clusters with low data density are identified, and the values ​​of the explanatory variables of the supplemented data are calculated so that data can be easily supplemented to the clusters with low data density.

[0038] (B) Clustering by the value of the objective variable is used when supplementing a target data set for learning a model so that a model can be created that accurately represents the relationship between the values ​​of multiple explanatory variables of actual measured data and the values ​​of explanatory variables in a portion of the data space of the actual measured data where the value of at least one objective variable is close to the target value (for example, a portion where a singular point such as a peak exists). (B) Clustering by the value of the objective variable identifies a cluster that includes actual measured data where the value of at least one objective variable is close to the target value, and calculates the values ​​of the explanatory variables of the supplemented data so that data can be easily supplemented to the cluster that includes actual measured data close to the target value.

[0039] For example, data may be supplemented to already obtained actual measurement data to verify whether the measurement target has a target physical property (objective variable), whether the measurement target has a better physical property (objective variable) than the target physical property (target value of the object variable), or whether the composition of the components of the measurement target or the manufacturing conditions can be changed as explanatory variables so as to have the target physical property (objective variable).In this case, the dataset expansion support device 1 performs (B) clustering based on the value of the object variable, making it easy to appropriately supplement the actual measurement data with data.

[0040] The model selection button 303 is a button for selecting the type of model (measured data model) that represents the relationship between the objective variable of the measured data to be analyzed and the explanatory variables in the dataset expansion support process, which will be described later with reference to Fig. 4. When the model selection button 303 is pressed, a list of model types is displayed, allowing the type of model to be selected.

[0041] The explanatory variable setting value input table 310 is a table used to input the settable range and measurable range for the explanatory variable. The explanatory variable setting value input table 310 has an explanatory variable number column 311, a settable range column 312 used to input the settable range, and a measurable range column 313 used to input the measurable range. The settable range and measurable range for one explanatory variable can be input in one row. That is, the user can input the settable range for each explanatory variable in a cell in the settable range column 312 in one row of the explanatory variable setting value input table 310. In addition, the user can input the measurable range for each explanatory variable in a cell in the measurable range column 313 in one row of the explanatory variable setting value input table 310. The settable range is the range of explanatory variables of actual measurement data that can be generated by a device (measuring instrument, etc.) that generates values ​​of explanatory variables of actual measurement data, and the range of explanatory variables of actual measurement data that can be set in the process of generating the actual measurement data. The measurable range is the range of explanatory variables of actual measured data that are capable of generating values ​​of explanatory variables of actual measured data (e.g., measuring equipment) in which the values ​​of explanatory variables of the actual measured data are significant, and the range of explanatory variables of actual measured data that can be set in the process of generating the actual measured data.

[0042] For example, in the case of a semiconductor manufacturing device using a plasma process, the measurable range is the setting range of adjustable parameters such as the adjustable inflow rate of process gas and the amount of power supply for generating plasma.

[0043] Here, if the inflow of the process gas and the power supply for plasma generation are not within an appropriate range, the plasma to be measured will not be generated. In the case of a semiconductor manufacturing device using a plasma process, the measurable range is the range of each parameter in which plasma can be generated, using parameters such as the inflow of the process gas and the amount of power supply for plasma generation.

[0044] If the settable range value is not set (input), it is assumed that there is no limit to the settable range, and the settable range is all values.Similarly, if the measurable range value is not set (input), it is assumed that there is no limit to the measurable range, and the measurable range is all values.

[0045] The delete row button 315 is a button that, when pressed, deletes the row next to the delete row button 314 in the explanatory variable setting value input table 310 .

[0046] The add row button 316 is a button that, when pressed, adds a row to the explanatory variable setting value input table 310 .

[0047] The objective variable setting value input table 320 is a table used for inputting target values ​​for objective variables. The objective variable setting value input table 320 has a column 321 for objective variable numbers and a column 322 for target values ​​used for inputting target values. The user can input a target value for one objective variable in a cell in the target value column 322 of one row of the objective variable setting value input table 320. A target value is a desired value of an explanatory variable. A target value is set for each explanatory variable. Values ​​such as extreme values ​​(maximum and minimum values) and 0 may be set by default as target values ​​in the objective variable setting value input table 320.

[0048] The delete row button 325 is a button that, when pressed, deletes the row next to the delete row button 325 from the objective variable set value input table 320. The add row button 326 is a button that, when pressed, adds a row to the objective variable set value input table 320.

[0049] When the location where the dataset to be analyzed (input dataset) is saved is entered in the input dataset save location input field 301, the dataset to be analyzed (input dataset) may be automatically analyzed, and rows in the explanatory variable setting value input table 310 and the objective variable setting value input table 320 may be automatically created so as to correspond to the actual measurement data to be analyzed.

[0050] When the send button 330 is pressed, dataset expansion information including the dataset to be analyzed (input dataset) stored in the location entered in the input dataset storage location input field 301, the clustering method specified in the clustering method selection button 302 (referred to as "clustering method specific information"), the type of model specified in the model selection button 303, the settable range and measurable range entered in the explanatory variable setting value input table 310, the target value of the objective variable entered in the objective variable setting value input table 320, and a command to execute dataset expansion support processing is sent to the dataset expansion support device 1.

[0051] <Dataset Expansion Support Processing> 4 is a flowchart showing an example of a dataset expansion support process according to an embodiment. As described above, the user operates the user terminal 500 to input dataset expansion information to be used in the dataset expansion process on the condition input screen (see FIG. 3) displayed on the user terminal 500, and causes the user terminal 500 to transmit the dataset expansion information to the dataset expansion support device 1. When the dataset expansion support device 1 receives the dataset expansion information, the dataset expansion support unit 122 executes the dataset expansion support process.

[0052] The target dataset is a dataset to be analyzed that is generated by extracting data from the input dataset. The target dataset includes a plurality of actual measurement data. As described above, the actual measurement data consists of a set of one or more objective variable values ​​and one or more explanatory variable values.

[0053] The dataset expansion support unit 122 extracts actual measurement data within the settable range and the measurable range from the input dataset included in the dataset expansion information to create a target dataset, acquires a pre-learning actual measurement data model corresponding to the actual measurement data from the actual measurement model store 231, and trains the acquired pre-learning actual measurement data model using the target dataset to generate an actual measurement data model (step S110). The dataset expansion support unit 122 extracts actual measurement data within the settable range and the measurable range from the input dataset included in the dataset expansion information to create a target dataset. As a result, all actual measurement data included in the target dataset has explanatory variables within the settable range and explanatory variables within the measurable range. Therefore, the actual measurement data included in the target dataset is data obtained under appropriately set conditions.

[0054] The dataset expansion support unit 122 acquires the number of explanatory variables and the number of target variables of the actual measurement data from the target dataset included in the dataset expansion information. The dataset expansion support unit 122 acquires the number of explanatory variables of the actual measurement data and a model corresponding to the number of explanatory variables from the actual measurement model store 231, and stores them as a pre-learned actual measurement data model. The dataset expansion support unit 122 trains the actual measurement data included in the target dataset (actual measurement data in which the values ​​of the explanatory variables are within a settable range and the values ​​of the explanatory variables are within a measurable range) into the pre-learned actual measurement data model, and stores the trained pre-learned actual measurement data model as a measured data model.

[0055] Next, the dataset expansion support unit 122 clusters the target dataset into a plurality of clusters (step S120). The clustering is performed by a method indicated in the clustering specific information included in the dataset expansion information. In the example of the condition input screen 300 shown in Fig. 3, two types of clustering specific information are given: (A) clustering based on data density, and (B) clustering based on the value of the objective variable. Below, examples of each of these two types of clustering methods will be described.

[0056] (A) Clustering of data by density. A1A calculates the distance (referred to as the "neighboring point distance dn") between each data point of the actual measured data in the data space of the explanatory variables and the closest data point, and stores the coordinates of the data point and the neighboring point distance dn in correspondence with each other. Here, the distance is the Euclidean distance. Mahalanobis distance or the like may also be used as the distance. The data space of the explanatory variables is a space in which the explanatory variables are used as coordinates. A data point is a point in the data space of the explanatory variables that represents the actual measured data and has the explanatory variables of the actual measured data as its coordinates. The data points are associated with the values ​​of the explanatory variables of the actual measured data, the values ​​of the objective variables, and the neighboring point distance dn. A1B, a data point whose neighboring point distance dn is equal to or less than the neighboring point distance threshold tdn (neighboring point distance dn ≦ neighboring point distance threshold tdn) is set as a candidate data point. A candidate data point is one that is located at a distance equal to or less than the neighboring point distance threshold tdn. Therefore, the candidate data point can be considered to exist in an area of ​​high data point density in the data space. A1C, if the distance between two of the multiple candidate data points is equal to or less than the group distance threshold tdg, the two candidate data points are grouped into one cluster. In this manner, the multiple candidate data points are classified into clusters by repeatedly grouping the candidate data points whose distance to neighboring candidate data points is equal to or less than the group distance threshold tdg into one cluster. The distance between the clusters is greater than the group distance threshold tdg. The group distance threshold tdg is a predetermined value that is greater than the neighboring point distance threshold tdn (neighboring point distance threshold tdn<group distance threshold tdg). A1D, in a cluster, a circle with a radius of the group distance threshold tdg is drawn with each candidate data point as its center (called a "candidate data circle"; if the data space of the explanatory variables is not two-dimensional, a line with a distance of the group distance threshold tdg is drawn with each candidate data point as its center), and the region in which the candidate data circle exists is defined as the cluster region. The area of ​​the cluster region (if the data space of the explanatory variables is not two-dimensional, volume or the like is used instead of area) is defined as the cluster area. The density of the cluster is the number of candidate data points included in the cluster divided by the cluster area (cluster density = number of candidate data points included in the cluster / cluster area). A1E: For data points that are not included in any cluster, the neighboring point distance threshold tdn of A1B and the group threshold of A1C are increased by a predetermined value, and the processes of A1B, A1C, and A1D are repeated to complete the clustering.

[0057] (B) Clustering by the value of the objective variable. B1A calculates the distance (referred to as "neighboring point distance dn") between each data point of the measured data in the data space of the explanatory variables and the closest data point, and stores the coordinates of the data point (the value of each explanatory variable of the measured data), the value of each objective variable of the measured data, and the neighboring point distance dn in association with each other. Here, the data set expansion support unit 122 performs the same process as A1A described above. B1B: A data point of the actual measured data in which the value yi of the objective variable of the actual measured data is closer to the target value yca than the reference value yc is set as a candidate data point. The target value yca is included in the dataset enrichment information. In addition, it can be classified into the following three cases 1-3. 1. The target value yca is greater than the values ​​yi of the objective variable of all the actual measured data (the value yi of the objective variable of the actual measured data < the target value yca). 2. The target value yca is within the range of the values ​​yi of the objective variable of all the actual measured data (the minimum value ymin of the values ​​yi of the objective variable of the actual measured data ≦ the target value yca ≦ (The maximum value ymax of the objective variable value yi of the actual measured data.) 3. The target value yca is smaller than the objective variable values ​​yi of all the actual measured data (target value yca<the objective variable value yi of the actual measured data). In the case of 1 (value of the objective variable of the actual measured data yi<target value yca), the data points of the actual measured data that satisfy the inequality "reference value yc≦value of the objective variable of the candidate data point yi≦maximum value ymax of the objective variable value yi" are set as candidate data points. In other words, the data points where the objective variable value yi of the actual measured data is equal to or greater than the reference value yc (reference value yc≦value of the objective variable of the candidate data point yi) are set as candidate data points. In the case of 2 (minimum value ymin of the objective variable value yi≦target value yca≦maximum value ymax of the objective variable value yi), the data points of the actual measured data that satisfy the inequality "reference value yc≦objective variable value yi of the candidate data point≦maximum value ymax of the objective variable value yi" are set as candidate data points. That is, the data points where the objective variable value yi of the actual measured data is equal to or greater than the reference value yc (reference value yc≦objective variable value yi of the candidate data point) are set as candidate data points. A second reference value yc2 may be set, and the data points where the objective variable value yi of the actual measured data is equal to or greater than the reference value yc and equal to or less than the second reference value yc2 (reference value yc≦objective variable value yi of the candidate data point≦second reference value yc2) may be set as candidate data points. In the case of 3 (target value yca<value yi of the objective variable of the actual measured data), the data points of the actual measured data that satisfy the inequality "minimum value ymin of the objective variable value yi ≦ value yi of the objective variable of the candidate data point ≦ maximum value ymin of the objective variable value yi + tolerance ycb" are selected as candidate data points. The reference value yc, the second reference value yc2, and the allowable value ycb are set for each objective variable. For example, the reference value yc in the above cases 1 and 2 may be the average value of the objective variable values ​​yi. Also, the allowable value ycb in the above case 3 may be the average value of the objective variable values ​​yi. In addition, when there are multiple objective variables of the actual measurement data, among the candidate data points calculated for each objective variable, data points that are determined as candidate data points for all the objective variables are determined as candidate data points. Here, data points that are determined as candidate data points for any of the objective variables may be determined as candidate data points. B1C: If the distance between two of the multiple candidate data points is equal to or less than the group distance threshold tdg, the two candidate data points are grouped into one cluster. In this manner, the process of grouping candidate data points whose distance to neighboring candidate data points is equal to or less than the group distance threshold tdg into one cluster is repeated to classify multiple measured data into clusters. The group distance threshold tdg is a predetermined value. B1D: In the cluster, a circle with a radius of the group distance threshold tdg is drawn with each candidate data point as the center (referred to as a "candidate data circle"; if the data space of the explanatory variables is not two-dimensional, a line with a distance of the group distance threshold tdg with each candidate data point as the center) is drawn, and the area where the candidate data circle exists is determined as the cluster area. B1E: For data points not included in any cluster, the reference value yc and tolerance value ycb of B1B and the group distance threshold tdg of B1C are increased by a predetermined value, and the processes of B1B, B1C, and B1D are repeated to complete the clustering.

[0058] Next, the dataset expansion support unit 122 selects a cluster to be supplemented with supplementary data from among the multiple clusters (step S130). Below, examples of a method for selecting a cluster to be supplemented will be described for each of the two types of clustering: (A) clustering based on data density and (B) clustering based on the value of a target variable.

[0059] (A) Clustering of data by density. The dataset expansion support unit 122 sets the number of clusters to be replenished (a predetermined number of clusters to be replenished) as the clusters to be replenished, in ascending order of cluster density, starting from the cluster with the smallest cluster density. The number of clusters to be replenished is a number equal to or greater than 1. If it is not necessary to replenish the cluster with the smallest cluster density with actual measurement data, the number of clusters to be replenished (a predetermined number of clusters to be replenished) may be set as the clusters to be replenished, in ascending order of cluster density, starting from the cluster with the smallest cluster density, among the clusters with cluster densities greater than a predetermined value (for example, the value of the cluster density for an area in the data space of the explanatory variables where the density of the actual measurement data is low because it can be determined that the actual measurement data is not significant).

[0060] (B) Clustering by the value of the objective variable. The dataset expansion support unit 122 calculates the maximum value ymax of the objective variable value yi of the actual measurement data among the data points included in the cluster for each cluster. Then, the dataset expansion support unit 122 selects the number of clusters to be replenished (a predetermined number of clusters to be replenished) as clusters to be replenished from the cluster with the largest maximum value ymax of the objective variable value yi of the actual measurement data. The number of clusters to be replenished is a number equal to or greater than 1.

[0061] Next, the dataset expansion support unit 122 uses the actual measurement data model to calculate a set of explanatory variable values ​​of one or more pieces of supplementary data to be supplemented to the supplement target cluster (all explanatory variable values ​​for one piece of supplementary data), and outputs the calculated set of explanatory variable values ​​of the supplementary data to the user terminal 500 (step S140). Below, examples of methods for calculating explanatory variable values ​​of supplementary data in each of two cases: (A) clustering by data density and (B) clustering by target variable value will be described.

[0062] (A) Clustering of data by density. The dataset expansion support unit 122 calculates a data point (referred to as a "theoretical replenishment point") in the cluster area of ​​each cluster to be replenished, where the value yi of the objective variable of the actual measurement data, which is the theoretical value calculated by the actual measurement data model, is closest to the target value. The target value is a value included in the dataset expansion information transmitted from the user terminal 500 (the target value input to the objective variable setting value input table 320 in FIG. 3). The value of the explanatory variable of the theoretical replenishment point is set as the value of the explanatory variable of the replenishment data.

[0063] (B) Clustering by the value of the objective variable. The dataset expansion support unit 122 calculates the following three data points (referred to as "theoretical replenishment points") where the value yi of the objective variable of the actual measurement data, which is the theoretical value calculated by the actual measurement data model, is closest to the target value. That is, 1. The theoretical replenishment point (referred to as "maximum theoretical replenishment point") where the value yi of the objective variable of the actual measurement data is closest to the target value yca. 2. The theoretical replenishment point (referred to as "minimum theoretical replenishment point") where the value yi of the objective variable of the actual measurement data is the farthest from the target value yca. The theoretical replenishment point (referred to as "intermediate theoretical replenishment point") where the value yi of the objective variable of the actual measurement data is the average of the value yi of the objective variable of the maximum theoretical replenishment point and the value yi of the objective variable of the minimum theoretical replenishment point. Then, at each of the maximum theoretical replenishment point, the minimum theoretical replenishment point, and the intermediate theoretical replenishment point, the values ​​of the explanatory variables are set to the values ​​of the explanatory variables of the replenishment data. In this way, for each cluster, we consider supplementing with one or more of the three pieces of supplementary data corresponding to the highest theoretical supplementary point, the lowest theoretical supplementary point, and the intermediate theoretical supplementary point.

[0064] Next, the dataset expansion support unit 122 waits for reception of at least one piece of supplementary data from the user terminal 500 (step S150). When the dataset expansion support device 1 receives the supplementary data from the user terminal 500, it executes the process of step S160.

[0065] Next, the dataset expansion support unit 122 generates a supplemented target dataset by adding at least one piece of supplementary data received from the user terminal 500 to the target dataset (step S160). Here, the dataset expansion support device 1 receives actual measurement data other than the actual measurement data (supplementary data) in which the set of explanatory variable values ​​of the supplementary data calculated in step S140 is used as explanatory variables together with the supplementary data, and in step S160, the dataset expansion support unit 122 may generate a supplemented target dataset by adding the received supplementary data and the actual measurement data other than the supplementary data to the target dataset.

[0066] Next, the dataset expansion support unit 122 trains the pre-learning actual data model with the replenished target dataset to generate an actual data model (step S170). Here, similar to step S110, the dataset expansion support unit 122 acquires a pre-learning actual data model corresponding to the number of objective variables and the number of explanatory variables of the actual data from the actual measurement model store 231, trains the acquired pre-learning actual data model with the replenished target dataset, and generates a replenished actual data model. The dataset expansion support unit 122 may also generate a model by additionally training the replenished data on the actual data model trained in S110, and use the generated model as the replenished actual data model.

[0067] Next, the dataset augmentation support unit 122 determines whether the actual measurement data model can correctly calculate the value of the objective variable of the supplementary data from the values ​​of the explanatory variables of the supplementary data (step S180). If the dataset augmentation support unit 122 determines that the actual measurement data model can correctly calculate the value of the objective variable of the supplementary data from the values ​​of the explanatory variables of the supplementary data (step S180: Yes), the process proceeds to step S190, and if the dataset augmentation support unit 122 determines that the actual measurement data model cannot correctly calculate the value of the objective variable of the supplementary data from the values ​​of the explanatory variables of the supplementary data (step S180: No), the process returns to step S120.

[0068] In both cases (A) of clustering by data density and (B) of clustering by the value of the objective variable, the dataset expansion support unit 122 determines whether or not "the actual measurement data model can correctly calculate the value of the objective variable of the supplementary data from the values ​​of the explanatory variables of the supplementary data" as follows.

[0069] The dataset expansion support unit 122 inputs the values ​​of the explanatory variables of the supplementary data for each piece of supplementary data, and causes the post-supplementation actual measurement data model to output the value of the objective variable of the supplementary data (referred to as the “theoretical objective variable value Yr”).

[0070] Furthermore, for each piece of supplementary data, if the difference dYre between the theoretical value Yr of the objective variable and the target value Ye of the objective variable of the supplementary data (difference dYre = theoretical value Yr of the objective variable - target value Ye of the objective variable of the supplementary data) is smaller than the target value Y of the supplementary data multiplied by a predetermined rate Rj (e.g., 0.1) (difference dYre = theoretical value Yr of the objective variable - target value Ye of the objective variable of the supplementary data < target value Ye of the objective variable of the supplementary data x predetermined rate Rj), then the judgment value J for the supplementary data is set to 1 (judgment value J = 1). On the other hand, when the difference dYre between the theoretical objective variable Yr and the target value Ye of the objective variable of the supplementary data (difference dYre=theoretical objective variable Yr-target value Ye of the objective variable of the supplementary data) is equal to or greater than the value obtained by multiplying the target value Y of the supplementary data by a predetermined rate Rj (e.g., 0.1) (difference dYre=theoretical objective variable Yr-target value Ye of the objective variable of the supplementary data≧target value Ye of the objective variable of the supplementary data x predetermined rate Rj), the judgment value J for the supplementary data is set to 0 (judgment value J=0). Here, the judgment value J is calculated based on the product of the target value Ye of the objective variable of the supplementary data and the predetermined rate Rj, as a relative value of the target value Ye of the objective variable of the supplementary data, and the judgment value J is calculated in consideration of the magnitude of the target value Ye of the objective variable of the supplementary data, thereby indicating whether the actual measurement data model can correctly calculate the target value Ye of the objective variable of the supplementary data. When there are multiple objective variables, the judgment value J is calculated for each of the objective variables as described above, and the average of the judgment values ​​J of the objective variables is set as the judgment value J for the supplementary data.

[0071] If the average value of the judgment value J for the supplementary data is greater than 0.5, the supplemented target dataset is deemed not to have sufficient quality and quantity of actual measurement data to create an appropriate actual measurement data model. That is, the dataset augmentation support unit 122 calculates the average value Jav of the judgment values ​​J for all supplementary data. If the average value Jav is less than 0.5, the dataset augmentation support unit 122 determines that "the actual measurement data model can correctly calculate the value of the objective variable of the supplementary data from the value of the explanatory variable of the supplementary data" (step S180: Yes). On the other hand, if the average value Jav is greater than 0.5, the dataset augmentation support unit 122 determines that "the actual measurement data model cannot correctly calculate the value of the objective variable of the supplementary data from the value of the explanatory variable of the supplementary data" (step S180: No).

[0072] Next, the dataset expansion support unit 122 outputs a message to the user terminal 500 stating that "the supplemented target dataset contains sufficient actual measurement data," and ends the process.

[0073] The process of step S120 is a clustering process for clustering the target data set into a plurality of clusters. The process of S130 is a replenishment target cluster selection process for selecting a replenishment target cluster to be replenished with replenishment data from a plurality of clusters. The process of step S140 includes an explanatory variable calculation process for calculating explanatory variable values ​​of the replenishment data based on an actual measurement data model that is trained using the target data set and represents the relationship between one or more explanatory variable values ​​and one or more objective variable values ​​of the actual measurement data, and at least one selected replenishment target cluster. The process of step S140 includes an explanatory variable output process for outputting the calculated explanatory variable values ​​of the replenishment data.

[0074] The process of step S170 is a verification actual measurement model learning process that generates a trained post-replenishment actual measurement data model using a post-replenishment target dataset in which the supplementary data is added to the target dataset. The process of S180 is a dataset determination process that evaluates the post-replenishment target dataset using a post-replenishment actual measurement data model trained on the actual measurement data model using the post-replenishment target dataset. The dataset determination process of S180 determines whether or not the value of the objective variable of the supplementary data can be correctly calculated from the value of the explanatory variable of the supplementary data using the post-replenishment actual measurement data model (step S180), and if it is determined that the value of the objective variable of the supplementary data cannot be correctly calculated from the value of the explanatory variable of the supplementary data using the post-replenishment actual measurement data model (step S180: No), a clustering process (step S120), a supplement target cluster selection process (step S130), an explanatory variable calculation process (step S140), and an explanatory variable output process (step S140) are executed to output the value of the explanatory variable of the supplementary data (step S140).

[0075] <Effects of the Invention> As described above, in the embodiment, the dataset expansion support device 1 (dataset expansion support method, dataset expansion support program) performs multiple clustering on the target dataset to be analyzed, thereby generating clusters that represent the collective properties of the multiple actual measurement data included in the clusters. This allows the dataset expansion support device 1 to support supplementing the target dataset with supplementary data, taking into account the collective properties of the multiple actual measurement data.

[0076] Then, by training the target dataset on the measured data model, it becomes possible to calculate the values ​​of the explanatory variables of the supplemental data using the measured data model including the properties of the target dataset. Therefore, the dataset expansion support device 1 (dataset expansion support method, dataset expansion support program) makes it easy to appropriately supplement data to the dataset used for model training.

[0077] In addition, the dataset expansion support device 1 (dataset expansion support method, dataset expansion support program) can reduce the energy required to replenish data and appropriately generate a dataset, as well as the amount of carbon dioxide emissions generated, thereby curbing global warming.

[0078] In addition, one type of clustering process clusters the target dataset into multiple clusters based on the density of the actual measured data in the data space of the explanatory variables of the actual measured data of the target dataset. This makes it possible to calculate a portion (cluster) to be supplemented with supplementary data in the data space of the explanatory variables of the actual measured data in order to create an actual measured data model that represents the relationship between the values ​​of multiple explanatory variables of the actual measured data and the values ​​of the explanatory variables throughout the data space of the explanatory variables of the actual measured data of the target dataset. As a result, it becomes easy to supplement the supplementary data to create a sufficient dataset in order to create an actual measured data model that represents the relationship between the values ​​of one or more explanatory variables of the actual measured data and the values ​​of one or more objective variables throughout the data space of the explanatory variables of the actual measured data of the target dataset.

[0079] In addition, by selecting a cluster that does not have the highest density of actual measured data as a cluster to be supplemented with data, a cluster that is lacking in data due to a low density of actual measured data can be appropriately selected as a cluster to be supplemented with data. This makes it easy to supplement with supplementary data to create a sufficient dataset in order to create a measured data model that represents the relationship between the values ​​of one or more explanatory variables of the measured data and the values ​​of one or more objective variables over the entire data space of the explanatory variables of the measured data of the target dataset.

[0080] Furthermore, the dataset expansion support device 1 (dataset expansion support method, dataset expansion support program) executes a verification measurement model learning process that generates an actual measurement model trained on an actual measurement data model using a supplemented target dataset in which supplementary data is added to the target dataset, and a dataset judgment process that evaluates the supplemented target dataset using an actual measurement model trained on the actual measurement data model using the supplemented target dataset. This allows the dataset expansion support device 1 (dataset expansion support method, dataset expansion support program) to reliably and appropriately supplement data to the dataset used for model training.

[0081] Furthermore, the dataset determination process (S180) determines whether the value of the objective variable of the supplementary data can be correctly calculated from the values ​​of the explanatory variables of the supplementary data using the actual measurement data model (S180), and if it is determined that the value of the objective variable of the supplementary data cannot be correctly calculated from the values ​​of the explanatory variables of the supplementary data using the actual measurement data model (S180: No), the clustering process (S120), the supplement target cluster selection process (S130), the explanatory variable calculation process (S140), and the explanatory variable output process (S140) are executed to output the explanatory variables of the supplementary data. This allows the dataset expansion support device 1 (dataset expansion support method, dataset expansion support program) to reliably and appropriately supplement data to the dataset used for model learning.

[0082] In addition, in one type of clustering process, in the case of (B) clustering by the value of the objective variable, the target dataset is clustered into multiple clusters based on the value of the objective variable of the actual measurement data. Therefore, from the viewpoint of the target value of the objective variable, the actual measurement data of the target dataset can be classified into clusters based on the properties of a group (cluster) of multiple actual measurement data. In this way, the dataset expansion support device 1 (dataset expansion support method, dataset expansion support program) makes it easy to supplement the target dataset with supplementary data so that actual measurement data can be obtained in which the value of the objective variable of the actual measurement data of the target dataset is close to the target value.

[0083] In addition, in the case of (B) clustering by the value of the objective variable, the cluster to be supplemented includes a cluster including actual measurement data in which the value of the objective variable is closest to the target value among the actual measurement data. This allows the dataset expansion support device 1 (dataset expansion support method, dataset expansion support program) to appropriately select a cluster to be supplemented.

[0084] For example, data may be supplemented to already obtained actual measurement data to verify whether the measurement target has a target physical property (objective variable), whether the measurement target has a better physical property (objective variable) than the target physical property (target value of the objective variable), or whether the composition of the components of the measurement target or the manufacturing conditions can be changed as explanatory variables so as to have the target physical property (objective variable).In this case, the dataset expansion support device 1 (dataset expansion support method, dataset expansion support program) performs (B) clustering by the value of the objective variable, making it easy to appropriately supplement the actual measurement data with data.

[0085] Furthermore, the dataset expansion support device 1 (dataset expansion support method, dataset expansion support program) calculates values ​​of explanatory variables for multiple supplementary data, thereby enabling the dataset expansion support device 1 (dataset expansion support method, dataset expansion support program) to support more appropriately supplementing data to a dataset used for model training.

[0086] Furthermore, the dataset expansion support device 1 (dataset expansion support method, dataset expansion support program) limits the actual measurement data included in the target dataset to only actual measurement data in which the values ​​of the explanatory variables are within a settable range (S110, S160). This allows the dataset expansion support device 1 (dataset expansion support method, dataset expansion support program) to support more appropriately supplementing the dataset used for model learning with data.

[0087] In addition, the dataset expansion support device 1 (dataset expansion support method, dataset expansion support program) limits the actual measurement data included in the target dataset to only actual measurement data in which the values ​​of the explanatory variables are within a measurable range (S110, S160).This allows the dataset expansion support device 1 (dataset expansion support method, dataset expansion support program) to assist in more appropriately supplementing the dataset used for model training with data.

[0088] In addition, each of the above configurations, functions, processing units, processing means, etc. may be realized in part or in whole by hardware, for example, by designing them as integrated circuits. The present invention can also be realized by software program code that realizes the functions of the embodiments. In this case, a storage medium on which the program code is recorded is provided to a computer, and a processor included in the computer reads the program code stored in the storage medium. In this case, the program code itself read from the storage medium realizes the functions of the above-mentioned embodiments, and the program code itself and the storage medium storing it constitute the present invention. Examples of storage media for supplying such program code include flexible disks, CD-ROMs, DVD-ROMs, hard disks, SSDs (Solid State Drives), optical disks, magneto-optical disks, CD-Rs, magnetic tapes, non-volatile memory cards, ROMs, etc.

[0089] Furthermore, the program code for realizing the functions described in this embodiment can be implemented in a wide range of program or script languages, such as assembler, C / C++, perl, Shell, PHP, Python, Java (registered trademark), etc.

[0090] Furthermore, the program code of the software that realizes the functions of the embodiments may be distributed over a network and stored in a storage means such as a computer's hard disk or memory, or in a storage medium such as a CD-RW or CD-R, and the processor of the computer may read out and execute the program code stored in the storage means or storage medium.

[0091] In the above-mentioned embodiment, the control lines and information lines are shown as those considered necessary for the explanation, and not all the control lines and information lines are shown in the product. All the components may be connected to each other. [Explanation of symbols]

[0092] 1: Data set expansion support device 11: Storage part 12: Processing section 121: Data response section 121a: Data response program 122: Data Set Expansion Support Department 122a;Dataset Expansion Support Program 21: Processor 22: Main memory 23: Secondary storage device 231: Actual Model Store 24: Input device 25: Output device 26: Network I / F 27: Bus 500: User terminal NW:Network

Claims

1. A dataset expansion support method in a dataset expansion support device comprising a processor and a storage device for storing information used in processing by the processor, the processor calculating information on supplementary data to be supplemented to an actual measurement dataset including a plurality of actual measurement data, each of which is a set of one or more target variable values ​​and one or more explanatory variable values, the method comprising: The processor, A clustering process for clustering the target dataset into multiple clusters; a supplementary cluster selection process for selecting at least one supplementary cluster to be supplemented with supplementary data from the plurality of clusters; an explanatory variable calculation process for calculating values ​​of explanatory variables of the supplementary data based on an actual measurement data model that is trained using the target dataset and that represents a relationship between values ​​of the one or more objective variables and values ​​of the one or more explanatory variables of the actual measurement data, and on at least one selected supplementary target cluster; and an explanatory variable output process for outputting the calculated explanatory variable values ​​of the supplementary data. How to help expand your dataset.

2. The method for supporting data set expansion according to claim 1, The clustering process clusters the target dataset into a plurality of clusters based on a density of the actual measurement data in a data space of the explanatory variables of the actual measurement data of the target dataset. How to help expand your dataset.

3. The method for supporting data set expansion according to claim 2, In the replenishment target cluster selection process, at least one of the replenishment target clusters is a cluster having a non-maximum density of the actual measurement data among the plurality of clusters. How to help expand your dataset.

4. The method for supporting data set expansion according to claim 1, Furthermore, when the supplementary data is acquired, A verification measurement model learning process for generating a trained supplemented measurement data model using a supplemented target data set in which the supplemented data is added to the target data set; Executing a dataset determination process in which the replenished target dataset is evaluated using a replenished actual data model trained on the actual data model, using the replenished target dataset; How to help expand your dataset.

5. The method for supporting data set expansion according to claim 4, further comprising the steps of: The data set determination process includes: determining whether or not a value of the objective variable of the supplementary data can be correctly calculated from a value of the explanatory variable of the supplementary data using the supplemented actual measurement data model; When it is determined that the value of the objective variable of the supplementary data cannot be correctly calculated from the values ​​of the explanatory variables of the supplementary data using the supplemented actual measurement data model, the clustering process, the supplement target cluster selection process, the explanatory variable calculation process, and the explanatory variable output process are executed to output the values ​​of the explanatory variables of the supplementary data. How to help expand your dataset.

6. The method for supporting data set expansion according to claim 1, The clustering process clusters the target data set into the plurality of clusters based on values ​​of the objective variable of the actual measurement data. How to help expand your dataset.

7. The method for supporting data set expansion according to claim 6, further comprising the steps of: At least one of the clusters to be replenished includes a cluster including actual measurement data in which the value of the objective variable is closest to a target value among the actual measurement data. How to help expand your dataset.

8. The method for supporting data set expansion according to claim 1, The number of the supplementary data is plural. How to help expand your dataset.

9. The method for supporting data set expansion according to claim 1, The actual measurement data included in the target data set is limited to only actual measurement data in which the value of the explanatory variable is within a settable range. How to help expand your dataset.

10. The method for supporting data set expansion according to claim 1, The actual measurement data included in the target data set is limited to only actual measurement data in which the value of the explanatory variable is within a measurable range. How to help expand your dataset.

11. A dataset expansion support device comprising: a processor; and a storage device for storing information used in processing by the processor, the processor calculating information regarding supplementary data to be supplemented to an actual measurement dataset including a plurality of actual measurement data each including a pair of values ​​of one or more objective variables and values ​​of one or more explanatory variables, The processor, A clustering process for clustering the target dataset into multiple clusters; a supplementary cluster selection process for selecting at least one supplementary cluster to be supplemented with supplementary data from the plurality of clusters; an explanatory variable calculation process for calculating values ​​of explanatory variables of the supplementary data based on an actual measurement data model that is trained using the target dataset and that represents a relationship between values ​​of the one or more objective variables and values ​​of the one or more explanatory variables of the actual measurement data, and on at least one selected supplementary target cluster; and an explanatory variable output process for outputting the calculated explanatory variable values ​​of the supplementary data. Data set expansion support device.

12. A dataset expansion support program for a dataset expansion support device, comprising: a processor; and a storage device for storing information used in processing by the processor, the processor calculating information regarding supplementary data to be supplemented to an actual measurement dataset including a plurality of actual measurement data, each of which is a set of one or more target variable values ​​and one or more explanatory variable values, The processor, A clustering process for clustering the target dataset into multiple clusters; a supplementary cluster selection process for selecting at least one supplementary cluster to be supplemented with supplementary data from the plurality of clusters; an explanatory variable calculation process for calculating values ​​of explanatory variables of the supplementary data based on an actual measurement data model that is trained using the target dataset and that represents a relationship between values ​​of the one or more objective variables and values ​​of the one or more explanatory variables of the actual measurement data, and on at least one selected supplementary target cluster; and an explanatory variable output process for outputting the calculated explanatory variable values ​​of the supplementary data. Data set expansion support program.

Images