Learning data selection method, learning data selection device, and computer-readable recording medium

Patent metadata

Field	Value
Publication number	US-11501203-B2
Application number	US-201816135446-A
Country	US
Kind code	B2
Filing date	Sep 19, 2018
Priority date	Sep 19, 2017
Publication date	Nov 15, 2022
Grant date	Nov 15, 2022

Abstract

A non-transitory computer-readable recording medium stores therein a learning data selection program that causes a computer to execute a process including: extracting a first input data group relating to first input data in correspondence with designation of the first input data included in an input data group input to a machine learning model, the machine learning model classifying or determining transformed data that is transformed from input data; acquiring a first transformed data group of the machine learning model and a first output data group of the machine learning model, respectively, the first transformed data group being input to the machine learning model and corresponding to the first input data group, the first output data group corresponding to the first transformed data group; and selecting learning target data of an estimation model from the first input data group.

First claim

What is claimed is:

1. A non-transitory computer-readable recording medium storing therein a learning data selection program that causes a computer to execute a process comprising:
extracting a first input data group relating to first input data in response to designation of the first input data included in an input data group input to a machine learning model, in accordance with a reference in which a collection timing of each data of the first input data group has a predetermined relationship with the first input data;
acquiring a first transformed data group generated by transforming the first input data group by the machine learning model, the first transformed data group including first transformed data based on the first input data;
calculating a first set of distances between the first input data and each data of the first input data group other than the first input data and a second set of distances between the first transformed data and each data of the first transformed data group other than the first transformed data;
selecting learning target data from the first input data group based on the first set of distances and the second set of distances; and
performing, based on the selected learning target data, machine learning of another machine learning model configured to estimate a factor which causes the machine learning model to output a classification result.

2. The non-transitory computer-readable recording medium according to claim 1 , wherein the learning target data is extracted based on data contents of each piece of output data included in the first output data group.

3. The non-transitory computer-readable recording medium according to claim 2 , wherein the learning target data is extracted based on a ratio of the data contents included in the first output data group.

4. The non-transitory computer-readable recording medium according to claim 3 , wherein the learning target data is extracted based on a ratio between a positive example and a negative example of the data contents included in the first output data group.

5. The non-transitory computer-readable recording medium according to claim 1 , wherein input data in which a data acquisition timing has a predetermined relationship with the first input data is specified as the first input data group.

6. The non-transitory computer-readable recording medium according to claim 1 , wherein input data in which a data generation source has a predetermined relationship with the first input data is specified as the first input data group.

7. The non-transitory computer-readable recording medium according to claim 1 , wherein the distance between the first input data and each piece of data of the first input data group is calculated by individually transforming the first input data and the data of the first input data group.

8. A learning data selection method comprising:
extracting a first input data group relating to first input data in response to designation of the first input data included in an input data group input to a machine learning model, in accordance with a reference in which a collection timing of each data of the first input data group has a predetermined relationship with the first input data;
acquiring a first transformed data group generated by transforming the first input data group by the machine learning model, the first transformed data group including first transformed data based on the first input data;
calculating a first set of distances between the first input data and each data of the first input data group other than the first input data and a second set of distances between the first transformed data and each data of the first transformed data group other than the first transformed data;
selecting learning target data from the first input data group based on the first set of distances and the second set of distances; and
performing, based on the selected learning target data, machine learning of another machine learning model configured to estimate a factor which causes the machine learning model to output a classification result, by a processor.

9. A learning data selection device comprising:
a processor configured to:
extract a first input data group relating to first input data in response to designation of the first input data included in an input data group input to a machine learning model, in accordance with a reference in which a collection timing of each data of the first input data group has a predetermined relationship with the first input data;
acquire a first transformed data group generated by transforming the first input data group by the machine learning model, the first transformed data group including first transformed data based on the first input data;
calculate a first set of distances between the first input data and each data of the first input data group other than the first input data and a second set of distances between the first transformed data and each data of the first transformed data group other than the first transformed data;
select learning target data from the first input data group based on the first set of distances and the second set of distances; and
perform, based on the selected learning target data, machine learning of another machine learning model configured to estimate a factor which causes the machine learning model to output a classification result.

Assignees

• Fujitsu Ltd

Inventors

• Goto Keisuke
• Maruhashi Koji
• Inakoshi Hiroya

Classifications

• G06F18/2133

  based on naturality criteria, e.g. with non-negative factorisation or negative correlation · CPC title

• G06F18/285

  Selection of pattern recognition techniques, e.g. of classifiers in a multi-classifier system · CPC title

• G06F18/24

  Classification techniques · CPC title

• G06F18/22 Primary

  Matching criteria, e.g. proximity measures · CPC title

• G06F18/214

  Generating training patterns; Bootstrap methods, e.g. bagging or boosting · CPC title