Performance Comparison of Machine Learning Algorithms for Imbalanced Class Classification in Hydraulic System

摘要

Several studies have been conducted over the past decades on fault detection for data-based system monitoring and have also been applied in various industry groups. However, there are various factors that make it difficult to apply in reality, even though algorithms for detecting various faults have been studied. The class imbalance problem is considered one of the most serious problems with fault detection. In many areas of classification, the data collected is largely imbalanced. Thus, in order to actually apply data-based system monitoring, there is a need for a method that can effectively solve such class imbalances. Various methodologies have been proposed to address this class imbalance problem, and good results have been recorded in several studies. In this paper, the overfitting problem caused by class imbalance is verified through experiment, and we find out which methodology is effective in solving class imbalance among several proposed methodologies for hydraulic system.