JP7626866B2 - Additional learning data selection device and computer-readable recording medium - Google Patents

Description

The present invention relates to an additional learning data selection device and a computer-readable recording medium.

At manufacturing sites such as factories, multiple industrial machines such as machine tools and robots are installed and operated to manufacture products. The operating status of the industrial machines is managed by analyzing data detected by sensors and the like. This data is analyzed by a control device that controls the industrial machines, or by a host computer connected to the control device via a wired or wireless network.

The operating status of industrial machinery may be estimated using machine learning technology. For example, a model for estimating the operating status of industrial machinery may be generated using machine learning technology, and the operating status of the industrial machinery may be estimated by applying data detected by the industrial machinery to the generated model. This estimation may include, for example, estimating whether the operating status is normal or abnormal, estimating the abnormal part, estimating the state of the components of the industrial machinery, etc. A specific example of estimating the state of the components of the industrial machinery is estimating the amount of thermal displacement. This is to estimate the amount of thermal displacement of each part of the industrial machinery due to changes in the ambient temperature of the industrial machinery and heat generated by the operation of the industrial machinery. The relationship between the temperature of each part of the industrial machinery and the amount of thermal displacement is modeled using machine learning, the temperature data of the industrial machinery when it is actually operating is detected, and the amount of thermal displacement of each part of the industrial machinery is estimated based on the detected data. The accuracy of machining can be improved by correcting the relative position of the workpiece and the tool based on the estimated amount of thermal displacement (for example, Patent Document 1, etc.).

When a model is created using machine learning technology, the validity of the model is verified. For the verification process, multiple validation data (hereafter referred to as a validation data group) are used, in which the relationship between data detected from industrial machinery (explanatory variables, also called input data or state data) and data indicating the phenomenon to be estimated (objective variables, also called output data or label data) is understood. The validation data group is applied to the model, and an estimated value is output. Then, it is verified whether the output estimated values each match the objective variables of the validation data, and the performance of the model is evaluated based on the results.

The verification data group used in the verification work is prepared in advance to satisfy the following requirements: (1) it covers all the patterns of explanatory variables expected from the machine specifications, etc., and (2) for all verification data, if the estimated value output by applying the model and the objective variable of the verification data are approximately the same, the performance of the model is guaranteed. For example, the verification data group used to verify the model for estimating the amount of thermal displacement described above needs to include verification data of a temperature pattern that is changed from the minimum temperature to the maximum temperature of the machine specifications, and to include verification data of a temperature pattern measured in an expected usage environment such as a user's factory.
A validation process is carried out using the validation data set prepared in this way. If the validation data is applied to the model and the output estimate does not approximately match the validation data's objective variable, this means that the performance of the created model is insufficient, and additional learning is carried out to recreate the model.

When performing additional learning, new pairs of explanatory variables and objective variables are prepared as additional learning data. The model is then updated by adding this additional learning data to the data used for the existing learning and learning it. After creating a new model, it is evaluated again using the same validation data, and this process is repeated until the performance of the model is guaranteed.

Patent No. 6753893

Here, if the additional training data used for additional training is inappropriate, the newly created model will not improve in performance compared to the model before additional training. On the other hand, when acquiring data to be used for training, comprehensive experiments over the range of explanatory variables can improve the performance of the model, but this requires a lot of effort and is inefficient.
Therefore, there is a need for technology that can determine in advance an appropriate set of explanatory variables that complement the data used for learning and efficiently prepare data to be used for additional learning.

The additional data selection device disclosed herein solves the above problem by selecting explanatory variables in the range that are under-learned in the data used for existing learning as data to be used for additional learning. To this end, the additional data selection device selects, from among a plurality of sets of explanatory variables presented as candidates, those having sets of explanatory variables that deviate greatly from the distribution of sets of explanatory variables in the data used for existing learning, as data to be used for additional learning.

One aspect of the present disclosure is an additional learning data selection device that creates a model based on machine learning of the relationship between one or more explanatory variables and a dependent variable and estimates a dependent variable corresponding to the explanatory variables using the model, the additional learning data selection device including: an additional learning candidate generation unit that generates a group of additional learning candidate data that are candidates for a set of explanatory variables to be used for additional learning that substantially covers a value space that a set of explanatory variables can take; a deviation calculation unit that calculates a deviation between a set of explanatory variables in learning data used to create the model and each of the additional learning candidate data included in the additional learning candidate data group ; an additional learning data selection unit that selects and displays data to be used for additional learning from the additional learning candidate data group based on the deviation; and a learning unit that performs additional learning of the model using the selected data and a dependent variable corresponding to the data obtained based on the selected data.

Another aspect of the present disclosure is a computer-readable recording medium having recorded thereon a program for operating a computer as an additional learning data selection device that creates a model based on machine learning of the relationship between one or more explanatory variables and a dependent variable and estimates a dependent variable corresponding to the explanatory variables using the model, the computer-readable recording medium having recorded thereon a program for operating a computer as an additional learning data selection device that creates an additional learning candidate generation unit that generates an additional learning candidate data group that is a candidate for a set of explanatory variables to be used for additional learning that substantially covers a value space that a set of explanatory variables can take; a deviation calculation unit that calculates a deviation between a set of explanatory variables in learning data used to create the model and each of the additional learning candidate data included in the additional learning candidate data group ; an additional learning data selection unit that selects and displays data to be used for additional learning from the additional learning candidate data group based on the deviation; and a learning unit that performs additional learning of the model using the selected data and a dependent variable corresponding to the data obtained based on the selected data.

One aspect of the present disclosure makes it possible to prepare data for additional learning that is expected to efficiently improve the performance of the model, thereby enabling additional learning to be performed efficiently.

FIG. 2 is a hardware configuration diagram of the additional learning data selection device. FIG. 2 is a block diagram showing functions of the additional learning data selection device. FIG. FIG. 13 is a diagram illustrating an example of additional learning candidate data. FIG. 13 is a plot of an example of additional learning candidate data. FIG. 13 is a plot of another example of additional learning candidate data. FIG. 13 is a plot of an example of a set of explanatory variables of learning data.

Hereinafter, an embodiment of the present invention will be described with reference to the drawings.
1 is a schematic hardware configuration diagram showing the main parts of an additional learning data selection device according to an embodiment of the present invention. The additional learning data selection device 1 of the present invention can be implemented as a control device that controls industrial machinery such as a machine tool based on a control program, for example. The additional learning data selection device 1 of the present invention can be implemented on a computer such as a personal computer attached to a control device that controls industrial machinery based on a control program, a personal computer connected to a control device via a wired/wireless network, a cell computer, a fog computer 6, or a cloud server 7. In this embodiment, an example is shown in which the additional learning data selection device 1 is implemented on a personal computer connected to a control device that controls industrial machinery via a network.

The CPU 11 provided in the additional learning data selection device 1 according to this embodiment is a processor that controls the entire additional learning data selection device 1. The CPU 11 reads out a system program stored in the ROM 12 via the bus 22, and controls the entire additional learning data selection device 1 in accordance with the system program. The RAM 13 temporarily stores temporary calculation data, display data, and various data input from outside.

The non-volatile memory 14 is composed of, for example, a battery-backed memory (not shown) or an SSD (Solid State Drive), and the memory state is maintained even when the power supply of the additional learning data selection device 1 is turned off. The non-volatile memory 14 stores data read from an external device 72 via the interface 15, data input via the input device 71, data acquired from the industrial machine 3 (including data detected by the sensor 4), and the like. The data stored in the non-volatile memory 14 may be expanded in the RAM 13 at the time of execution/use. In addition, various system programs such as a known analysis program are written in advance in the ROM 12.

Sensors 4 are attached to the industrial machine 3 to detect physical quantities such as current, voltage, vibration, and temperature of each part when the industrial machine 3 is operating. Examples of the industrial machine 3 include machining centers, lathes, and electric discharge machines. In response to a request from the additional learning data selection device 1, the industrial machine 3 transmits data such as the position, speed, acceleration, jerk of each axis during operation, the motor rotation speed, vibration, the temperature of each part, and the ambient temperature via the network 5.

The interface 15 is an interface for connecting the CPU 11 of the additional learning data selection device 1 to an external device 72 such as a USB device. For example, pre-stored control programs and data related to the operation of each industrial machine 3 can be read from the external device 72. In addition, the control programs and setting data edited within the additional learning data selection device 1 can be stored in an external storage means via the external device 72.

The interface 20 is an interface for connecting the CPU of the additional learning data selection device 1 to a wired or wireless network 5. The network 5 is connected to industrial machines 3, fog computers 6, cloud servers 7, etc., and exchanges data with the additional learning data selection device 1.

The display device 70 displays various data loaded into the memory, data obtained as a result of executing a program, data output from the machine learning device 100 (described later), and the like, which are output via the interface 17. The input device 71, which is composed of a keyboard, a pointing device, and the like, transmits commands and data based on operations by an operator to the CPU 11 via the interface 18.

The interface 21 is an interface for connecting the CPU 11 and the machine learning device 100. The machine learning device 100 includes a processor 101 that controls the entire machine learning device 100, a ROM 102 that stores system programs, etc., a RAM 103 for temporary storage in each process related to machine learning, and a non-volatile memory 104 used to store models, etc. The machine learning device 100 can observe each piece of information that can be acquired by the additional learning data selection device 1 via the interface 21 (for example, data detected during operation of the industrial machine 3). In addition, the additional learning data selection device 1 acquires the processing results output from the machine learning device 100 via the interface 21, and stores or displays the acquired results, or transmits them to other devices via the network 5, etc.

Figure 2 is a schematic block diagram showing the functions of the additional learning data selection device 1 according to the first embodiment of the present invention. Each function of the additional learning data selection device 1 according to this embodiment is realized by the CPU 11 of the additional learning data selection device 1 shown in Figure 1 and the processor 101 of the machine learning device 100 executing a system program and controlling the operation of each part of the additional learning data selection device 1 and the machine learning device 100.

The additional learning data selection device 1 of this embodiment includes a data acquisition unit 110, a data selection unit 120, a learning unit 130, an estimation unit 140, a verification unit 150, an additional learning candidate generation unit 160, a deviation calculation unit 170, and an additional learning data selection unit 180. In addition, on the RAM 13 to the non-volatile memory 14 of the additional learning data selection device 1, a data storage unit 210 which is an area for storing data acquired by the data acquisition unit 110, a learning data storage unit 220 and a verification data storage unit 230 which are areas for storing the learning data and the verification data selected by the data selection unit 120, respectively, are prepared in advance. In addition, on the RAM 103 to the non-volatile memory 104 of the machine learning device 100, a model storage unit 240 which is an area for storing a model created by the learning unit 130 through machine learning is prepared in advance.

The data acquisition unit 110 acquires data indicating the operating status of the industrial machine 3 detected by the sensor 4 etc. when the industrial machine 3 is in operation. The data acquired by the data acquisition unit 110 includes, for example, the position of each axis, speed, acceleration, jerk, motor rotation speed, vibration, temperature of each part, ambient temperature, etc., which indicate the operating status of the industrial machine 3. The data acquisition unit 110 may acquire data directly from the industrial machine 3 via the network 5, or may acquire data via an external device 72, fog computer 6, cloud server 7, etc. These data are mainly used as a set of explanatory variables in the model learning and verification operations.

The data acquisition unit 110 may also acquire data indicating the normal/abnormal state of the industrial machine 3, data related to the abnormal location, and data indicating the state of the components of the industrial machine. These data include the position, speed, acceleration, jerk, motor rotation speed, vibration, temperature of each part, and ambient temperature of each axis indicating the operating status of the industrial machine 3. The data acquisition unit 110 may acquire data directly from the industrial machine 3 via the network 5, or may acquire data via an external device 72, a fog computer 6, a cloud server 7, etc. In addition, data input by an operator from the input device 71 may be acquired for the purpose of labeling a set of predetermined explanatory variables. These data are mainly used as objective variables in the learning and verification work of the model.

The data acquisition unit 110 associates the data acquired to be used as a set of explanatory variables with the corresponding data acquired as the objective variable and stores them in the data storage unit 210.

The data selection unit 120 selects data that is a set of explanatory variables and objective variables stored in the data storage unit 210 into data to be used for learning the model and data to be used for verifying the model. For example, the data selection unit 120 may select a predetermined number of data into learning data, and select a predetermined number of data from the remaining data into a group of evaluation data. Alternatively, a predetermined selection rule may be determined in advance, and data for learning and data for verification may be selected according to the selection rule. As an example of the selection rule, a random number or the like may be used to determine the selection of each data, or a rule may be used that selects data so that the values of each explanatory variable are evenly distributed in order to prevent bias in the data selected for learning and verification to some extent. The data selection unit 120 stores the data selected for learning in the learning data storage unit 220, and the data selected for verification in the verification data storage unit 230.

The learning unit 130 performs machine learning using the learning data stored in the learning data storage unit 220. The learning unit 130 creates a model that learns the correlation between the set of explanatory variables indicated in the learning data and the objective variable by a known machine learning method, and stores the created model in the model storage unit 240. The machine learning method performed by the learning unit 130 may be, for example, a known method such as linear regression, a neural network, a multilayer perceptron, or a support vector machine.

The estimation unit 140 applies the set of explanatory variables input as the estimation target to the model created by the learning unit 130 and outputs an estimate of the objective variable.

The verification unit 150 verifies the performance of the model created by the learning unit 130 using the verification data stored in the verification data storage unit 230. The verification unit 150 outputs a set of explanatory variables of the verification data stored in the verification data storage unit 230 to the estimation unit 140 to obtain an estimate of the objective variable. Then, the verification unit 150 verifies whether the obtained estimate of the objective variable matches the objective variable of the verification data. If the error between the estimate of the objective variable estimated by the estimation unit 140 and the objective variable of the verification data is within a predetermined threshold range for all of the verification data stored in the verification data storage unit 230, the verification unit 150 evaluates that the performance of the model created by the learning unit 130 is excellent. Furthermore, if there is an error between the estimate of the objective variable estimated by the estimation unit 140 and the objective variable of the verification data for one or more verification data, which exceeds a predetermined threshold, the verification unit 150 evaluates that the performance of the model created by the learning unit 130 is poor. If the verification unit 150 evaluates that the performance of the model created by the learning unit 130 is poor, the additional training data selection device 1 decides to perform additional training.

The additional learning candidate generation unit 160 creates candidate data for additional learning when it is decided to perform additional learning. Candidate data for additional learning is data for which at least a set of explanatory variables is specified. The additional learning candidate generation unit 160 generates candidate data for additional learning, for example, by combining possible values for each explanatory variable.

3 to 6, the generation of additional learning candidate data by the additional learning candidate generator 160 will be described taking as an example the estimation of a thermal displacement compensation amount in a machine tool as the industrial machine 3.
FIG. 3 is a side view of a machine tool. In this example, a sensor 4a is attached to the base of the machine tool, a sensor 4b is attached to the column, and a sensor 4c is attached to the spindle, respectively, to detect the temperature of each part. The temperature detected by each sensor is transmitted to the additional learning data selection device 1 by the control device of the machine tool. This data is used as a set of explanatory variables. In addition, an operator measures the thermal displacement amount in the machine tool and inputs it to the additional learning data selection device 1. This data is used as a target variable. By repeating such experiments, a sufficient number of data (sets of explanatory variables and target variables) for generating and verifying a model is stored in the data storage unit 210 of the additional learning data selection device 1. This data is sorted by the data selection unit 120 into data for learning and data for verification. Then, the learning unit 130 performs machine learning based on the learning data, and a model that learns the correlation between the temperature of each part of the machine tool and the thermal displacement amount is created and stored in the model storage unit 240. Then, the verification unit 150 verifies the model using the verification data, and when it evaluates that the performance of the model is poor, it decides to perform additional learning.

FIG. 4 shows an example of additional learning candidate data generated by the additional learning candidate generating unit 160. In the example of FIG. 4, temperature A indicates the temperature detected by the sensor 4a, temperature B indicates the temperature detected by the sensor 4b, and temperature C indicates the temperature detected by the sensor 4c. As shown in the example of FIG. 4, the additional learning candidate generating unit 160 generates one or more sets of temperature A, temperature B, and temperature C as explanatory variables. The value of each explanatory variable may be a random number generated between the minimum and maximum values that the explanatory variable can take, and the value may be used. For example, if the minimum value of temperature A is TA _min and the maximum value is TA _max , a random number is generated in the range of TA _min to TA _max , and this is used as the explanatory variable value of temperature A. FIG. 5 shows an example of plotting additional learning candidate data generated using random numbers.

Furthermore, the additional learning candidate generation unit 160 may generate candidate data for additional learning so that the data is distributed approximately uniformly in a value space that the pair of explanatory variables can take. FIG. 6 shows an example of plotting additional learning candidate data generated so that the data is distributed approximately uniformly in a value space that the pair of explanatory variables can take. In this case, for example, values are generated for temperature A at a predetermined increment TA _div between TA _min and TA _max . Values are also generated similarly for temperature B and temperature C. Then, the generated values can be combined to generate a pair of explanatory variables.

The deviation calculation unit 170 calculates the deviation between the set of explanatory variable pairs in the learning data used to generate the model and each additional learning candidate data generated by the additional learning candidate generation unit 160. A known method such as cluster analysis or the MT method (Mahalanobis Taguchi Method) may be used to calculate the deviation. For example, when the MT method is used, the following formula 1 can be used to calculate the deviation d between the set of explanatory variable pairs in the learning data and the additional learning candidate data. In formula 1, x represents a vector that lists the explanatory variables of the additional learning candidate data, μ represents a vector that lists the average values of each explanatory variable in the learning data, and Σ represents a variance-covariance matrix for the explanatory variables in the learning data.

The additional learning data selection unit 180 selects data to be used for additional learning from the additional learning candidate data based on the deviation of each additional learning candidate data calculated by the deviation calculation unit 170. The additional learning data selection unit 180 may select the additional learning candidate data with the largest deviation as the data to be used for additional learning. Also, the additional learning data selection unit 180 may select a predetermined number of additional learning candidate data from the additional learning candidate data with the largest deviation as the data to be used for additional learning. Furthermore, the additional learning candidate data with a deviation degree equal to or greater than a predetermined threshold may be selected as the data to be used for additional learning.

Figure 7 shows an example of plotting sets of explanatory variables in learning data. In Figure 7, the white circles are plots of sets of explanatory variables in learning data. The black triangles and black circles are plots of additional learning candidate data. As is clear from Figure 7, the additional learning candidate data of the black triangles has a small deviation from the set of explanatory variable sets in the learning data. Since such data is unlikely to provide information beyond the range that could be learned with the existing learning data, it is considered that the learning effect is low even if it is used for additional learning. In contrast, the additional learning candidate data of the black circles has a large deviation from the set of explanatory variable sets in the learning data. Since such data is likely to provide information in a range that could not be learned with the existing learning data, the learning effect may be increased by using it for additional learning. Based on this idea, in the additional learning data selection device 1 according to this embodiment, additional learning candidate data with a high deviation is used for additional learning.

The data selected by the additional learning data selection unit 180 is displayed, for example, on the display device 70. Alternatively, it is transmitted to the industrial machine 3 via the network 5 and displayed on a display device (not shown) equipped on the industrial machine 3. The operator performs an experiment on the industrial machine 3 based on the displayed selected data, and obtains a target variable corresponding to the displayed data. The obtained target variable is then input to the additional learning data selection device 1, and learning data for additional learning is created in association with the data to be used for additional learning, and stored in the learning data storage unit 220. The learning unit 130 then performs additional learning using this data, updates the model, and stores it in the model storage unit 240.

The additional learning data selection device 1 having the above configuration can prepare data for additional learning that is expected to efficiently improve the performance of the model. The operator can efficiently collect data for additional learning by performing additional experiments only on the data presented by the additional learning data selection device 1. The effort of performing unnecessary experiments can be reduced. In this embodiment, the temperature of each part of the machine tool detected by the temperature sensor is set as an explanatory variable, and the thermal displacement amount of the machine tool in such a temperature distribution state is set as an objective variable, and a model that learns the correlation between them is created. Based on the estimated value of the thermal displacement amount by the created model, the thermal displacement is corrected in the machine tool. Learning such a model requires data that measures the thermal displacement amount in various temperature distributions, and the number of experiments is inevitably large. In addition, when creating a model, multiple additional experiments may be required until a good model can be created. The additional learning data selection device 1 according to this embodiment can significantly reduce the number of additional experiments by selecting appropriate data for additional learning.

As a modified example of the additional learning data selection device 1 according to the present embodiment, when data that has not been selected as learning data or verification data remains in the data storage unit 210, the additional learning candidate generation unit 160 may use a set of explanatory variables of the data as additional learning candidate data. Alternatively, a set of explanatory variables of verification data (verification data whose estimated value error exceeds a predetermined threshold) that caused the model to be determined to be defective among the verification data stored in the verification data storage unit 230 may be used as additional learning candidate data. Then, when the additional learning candidate data generated in this way is selected by the additional learning data selection unit 180 as data to be used for additional learning, the original data including the objective variable may be added to the learning data. By configuring in this way, when data suitable for additional learning already exists including the objective variable, additional learning can be performed on the data without performing additional experiments again, which also reduces the effort of performing unnecessary experiments.

Although the embodiments of the present invention have been described above, the present invention is not limited to the above-described examples of the embodiments, and can be embodied in various forms by making appropriate modifications.
For example, in the above embodiment, an example is shown in which a model is learned to estimate a thermal displacement amount as a response variable using temperature values detected at three locations on a machine tool as explanatory variables. However, the number of explanatory variables is not limited to three. For example, the present invention can be applied to cases in which a response variable is estimated using two explanatory variables or four or more explanatory variables.

Furthermore, when training a model to estimate the amount of thermal displacement, in addition to using the temperature values obtained by a temperature sensor, any data indicating conditions that affect the thermal displacement can be used as an explanatory variable. For example, it is known that heat generation in machine tools is affected by the motor's rotation speed and continuous rotation time. Therefore, it is possible to obtain data such as the operation patterns of the spindle motor and the feed axis motor and use them as explanatory variables to estimate the amount of thermal displacement.

When learning a model that estimates thermal displacement, in addition to the current temperature, the temperature value at a specified time in the past may be used as an explanatory variable. In this case, the temperature at the current time and the temperature at a past time may be combined and used as an explanatory variable.

In the above embodiment, an example was shown in which a model for estimating the amount of thermal displacement of a machine tool was used. However, the additional learning data selection device 1 of the present invention also operates suitably when other estimation models, such as estimation of normal or abnormal operating conditions, estimation of abnormal locations, and estimation of the state of components of industrial machinery, are used. For example, it can also be applied to the learning of a model for diagnosing motor faults. Such a model estimates the degree of abnormality indicating whether the motor is faulty from the measurement value of a sensor attached to the motor. As explanatory variables, the motor current value, motor rotation speed, torque command value, motor temperature, room temperature, and the value detected by a sensor of the sound generated by the motor can be used.

REFERENCE SIGNS LIST 1 Additional learning data selection device 3 Industrial machine 4 Sensor 5 Network 6 Fog computer 7 Cloud server 11 CPU
12 ROM
13 RAM
14 Non-volatile memory 15 Interface 17, 18, 20, 21 Interface 22 Bus 70 Display device 71 Input device 72 External device 100 Machine learning device 101 Processor 102 ROM
103 RAM
Reference Signs List 104 Non-volatile memory 110 Data acquisition unit 120 Data selection unit 130 Learning unit 140 Estimation unit 150 Verification unit 160 Additional learning candidate generation unit 170 Deviation calculation unit 180 Additional learning data selection unit 210 Data storage unit 220 Learning data storage unit 230 Verification data storage unit 240 Model storage unit

Claims (5)

1. An additional learning data selection device that creates a model based on machine learning of a relationship between one or more explanatory variables and a dependent variable, and estimates a dependent variable corresponding to the explanatory variable using the model,
an additional learning candidate generation unit that generates an additional learning candidate data group that is a candidate for a set of explanatory variables used in additional learning that substantially covers a value space that the set of explanatory variables can take ;
a deviation calculation unit that calculates a deviation between a set of explanatory variables in the learning data used to create the model and each of the additional learning candidate data included in the additional learning candidate data group ;
an additional learning data selection unit that selects and displays data to be used for additional learning from the additional learning candidate data group based on the deviation degree;
a learning unit that performs additional learning of the model by using the selected data and a corresponding objective variable obtained based on the selected data ;
An additional learning data selection device comprising: a thermal displacement amount of the industrial machinery is used as a response variable, and the thermal displacement amount of the industrial machinery is used as an explanatory variable; and the thermal displacement of the industrial machinery is corrected using the estimated thermal displacement amount of the machinery.
The additional learning data selection device according to claim 1 . The data indicating the state includes one or more of a machine temperature, a temperature around the machine, a spindle operation pattern, and a feed axis operation pattern.
The additional learning data selection device according to claim 2. The data indicating the state includes data acquired at a plurality of different times.
The additional learning data selection device according to claim 2. A computer-readable recording medium having recorded thereon a program for operating a computer as an additional learning data selection device that creates a model based on machine learning of the relationship between one or more explanatory variables and a dependent variable, and estimates a dependent variable corresponding to the explanatory variable using the model, the additional learning candidate generation unit generating an additional learning candidate data group that is a candidate for a set of explanatory variables used in additional learning that substantially covers a value space that a set of explanatory variables can take;
a deviation calculation unit that calculates a deviation between a set of explanatory variables in the learning data used to create the model and each of the additional learning candidate data included in the additional learning candidate data group ;
an additional learning data selection unit that selects and displays data to be used for additional learning from the additional learning candidate data group based on the deviation degree;
a learning unit that performs additional learning of the model by using the selected data and a target variable corresponding to the data obtained based on the selected data ;
A computer-readable recording medium having a program recorded thereon that causes a computer to function as a

Images