불균형 데이터 해소를 위한 오버샘플링 비교연구 = Over-Sampling Comparative Study for Imbalanced Data Classification

다국어 초록 (Multilingual Abstract)

Recently, as the problem of classifying and predicting imbalanced data has increased, research in various fields has intensified to solve it. This study compares re-sampling methods to address the classification problem of imbalanced data. Four classifiers, including the logistic regression model, were utilized, with AUC and F1-score serving as performance metrics in this study. The experiment yielded varied results based on the classifier or performance metrics used, and even when the classifier and performance metrics were identical, outcomes differed depending on the sample size or the imbalance rate. Therefore, this indicates the importance of making decisions based on comprehensive sensitivity analysis rather than relying solely on one data processing technique or classifier.