[2G4-GS-2f] 機械学習：データ分布

[2G4-GS-2f-01] Support Vector Machine considering the imbalance of data generation

〇Takuya Shimada¹, Takahiro Nishigaki¹, Takashi Onoda¹ (1. Aoyama Gakuin University)

Keywords:Machine Learning, Support Vector Machine, Mathematical Optimization

In recent years, machine learning has begun to be used for equipment abnormality diagnosis. In problems such as equipment abnormality diagnosis that classify normal or abnormal states, normal state data is likely to occur, and a large amount of data exists. On the other hand, abnormal state data rarely occurs, and only a very small amount of data exists. A support vector machine (SVM) that considers the imbalance between the amount of normal data and the amount of abnormal data has been proposed. However, the conventional method considers the imbalance of only the amount of data, and does not consider the imbalance of the data generation itself. In this paper, we propose an SVM that considers the imbalance of normal data and abnormal data generation itself. We also compare the conventional undersampling SVM and oversampling SVM to clarify the difference.