Machine learning device and machine learning method

1. A machine learning device, comprising:

a sparse modeling processing unit that acquires a plurality of individual importance degrees for each of a plurality of explanatory variable candidates, the plurality of individual importance degrees being acquired by using respective sparse modeling methods different from each other, each of the sparse modeling methods taking input data comprising a specified objective variable in a learning model used for industrial activity and the plurality of explanatory variable candidates that are candidates for an explanatory variable for explaining the specified objective variable; and

a selection unit that calculates a comprehensive importance degree for each of the plurality of explanatory variable candidates based on the plurality of individual importance degrees of each of the plurality of explanatory variable candidates, and selects an explanatory variable of the learning model from among the plurality of explanatory variable candidates based on the comprehensive importance degree.

2. The machine learning device according to claim 1 , further comprising:

an adjustment unit that adjusts the plurality of individual importance degrees in accordance with the respective sparse modeling processing methods, and calculates a plurality of adjusted individual importance degrees,

wherein the selection unit calculates the comprehensive importance degree based on the plurality of adjusted individual importance degrees.

3. The machine learning device according to claim 2 ,

wherein each of the individual importance degree is a contribution degree which is a value derived by each of the sparse modeling methods and which indicates a strength of a relationship between the specified objective variable and each of the plurality of explanatory variable candidates, and

wherein the adjustment unit extracts the explanatory variable candidate having the contribution degree equal to or greater than a predetermined threshold, and calculates the plurality of adjusted individual importance degrees for the extracted explanatory variable candidate.

4. The machine learning device according to claim 2 ,

wherein the adjustment unit extracts a predetermined number of the explanatory variable candidates having high individual importance degrees, and calculates the plurality of adjusted individual importance degrees for each of the extracted explanatory variable candidates.

5. The machine learning device according to claim 2 ,

wherein the adjustment unit calculates, as the adjusted individual importance degrees, scores corresponding to the plurality of individual importance degrees, and

wherein the selection unit calculates the comprehensive importance degree based on the scores.

6. The machine learning device according to claim 5 ,

wherein the adjustment unit calculates, as the adjusted individual importance degrees, the scores which are assigned in like an arithmetical series in accordance with the plurality of individual importance degrees.

7. The machine learning device according to claim 5 ,

wherein the adjustment unit calculates, as the adjusted individual importance degrees, the scores which are assigned in like a geometric series in accordance with the plurality of individual importance degrees.

8. The machine learning device according to claim 1 ,

wherein the selection unit calculates the comprehensive importance degree based on at least one of a sum, a product, and an average of the adjusted individual importance degrees for each of the plurality of explanatory variable candidates.

9. The machine learning device according to claim 1 ,

wherein the selection unit selects, as the explanatory variable, the explanatory variable candidate having the comprehensive importance degree equal to or greater than a predetermined threshold.

10. The machine learning device according to claim 1 ,

wherein the selection unit selects, as the explanatory variable, a predetermined number of explanatory variable candidates having high comprehensive importance degrees.

11. The machine learning device according to claim 1 , further comprising:

a learning model generation unit that generates the learning model by machine learning in which the explanatory variable candidate selected by the selection unit is used as learning data.

12. The machine learning device according to claim 1 , further comprising:

a learning model storage unit that stores the learning model generated by machine learning in which the explanatory variable selected by the selection unit is used as learning data; and

a prediction unit that performs a predetermined prediction, based on the explanatory variable and the learning model.

13. The machine learning device according to claim 12 ,

wherein the machine learning device is provided in a processing device that performs predetermined processing, or in a management device that performs overall management of a plurality of the processing devices, and

wherein the prediction unit predicts abnormality detection of the processing device, optimization of operation conditions of the processing device, or a processing result of the processing device.

14. The machine learning device according to claim 13 ,

wherein the processing device comprises:

an operation command unit that sends operation command data to a control unit; and

an operation command data adjustment unit that adjusts the operation command data based on a prediction result of the prediction unit.

15. The machine learning device according to claim 13 ,

the processing device is a molding machine, and

the prediction unit predicts a quality factor of a molded product to be produced by the molding machine.

16. A machine learning method, comprising:

acquiring a plurality of individual importance degrees for each of a plurality of explanatory variable candidates, the plurality of individual importance degrees being acquired by using respective sparse modeling methods different from each other, each of the sparse modeling methods taking input data comprising a specified objective variable in a learning model used for industrial activity and the plurality of explanatory variable candidates that are candidates for an explanatory variable for explaining the specified objective variable; and

calculating a comprehensive importance degree for each of the plurality of explanatory variable candidates based on the plurality of the individual importance degrees of each of the plurality of explanatory variable candidates, and selecting an explanatory variable of the learning model from among the plurality of explanatory variable candidates based on the comprehensive importance degree.

17. The machine learning method according to claim 16 , further comprising:

adjusting the plurality of individual importance degrees in accordance with the respective sparse modeling methods, and calculating a plurality of adjusted individual importance degrees; and

calculating the comprehensive importance degree based on the plurality of adjusted individual importance degrees.