GAMC: An Oversampling Method Based on Genetic Algorithm and Monte Carlo Method to Solve the Class Imbalance Issue in Industry
Xuekang Fan, Hong Zhi Yu · 2022
Due to the characteristics of long-term steady in industrial production processes, there is a large amount of imbalanced data in it, which can make it difficult for detection models to accurately capture abnormal events and cause property damage. In this paper, a data oversampling method is proposed to balance the industrial process data from the perspective of sample distribution. This oversampling method uses the crossover and mutation processes of genetic algorithm to generate offspring samples, and then uses the monte carlo acceptance-rejection sampling method to determine whether the new attribute values match the distribution of original sample attribute values. This method can effectively enrich the information carried in the minority class samples and improve the effectiveness of classification model. In the experimental section, the results of the comparison with existing two classical oversampling methods on six publicly available datasets show that the algorithm in this paper outperforms the existing oversampling methods in terms of overall performance.