2022 Fiscal Year Final Research Report
Data balancing for regression using imbalanced dataset
Project/Area Number |
21K21297
|
Research Category |
Grant-in-Aid for Research Activity Start-up
|
Allocation Type | Multi-year Fund |
Review Section |
1001:Information science, computer engineering, and related fields
|
Research Institution | Kyoto Tachibana University |
Principal Investigator |
|
Project Period (FY) |
2021-08-30 – 2023-03-31
|
Keywords | 機械学習 / 不均衡データ / データバランシング / 分類問題 / 回帰問題 |
Outline of Final Research Achievements |
We propose methods to address the imbalance of estimated values in regression and classification problems, respectively. The first method is a data balancing technique for regression problems using time series data as explanatory variables. This method generates new samples by interpolating time series data from two extracted samples in the dataset. Through performance evaluation, we found that it is possible to improve the estimation accuracy for minority data while suppressing the increase in mean absolute error. The second method is a data balancing technique for classification problems using conditional generative adversarial networks. Through performance evaluation using open datasets, we found that the proposed method achieved training a well-balanced estimator.
|
Free Research Field |
情報ネットワーク
|
Academic Significance and Societal Importance of the Research Achievements |
利用者が気づきにくい不均衡データによる推定値の偏りを軽減する手法を提案し,様々な機械学習との組み合わせ・応用を可能とする点が本研究の社会的意義である.特に近年ではセンシングデバイスの小型化・低価格化が進み,機械学習の科学・医療など様々な分野への応用手法が開発されていることから,今後ますますモバイル・ユビキタス分野において機械学習は利用されることが予想される.そのような応用事例において本研究は大きな役割を果たすと申請者は考える.
|