A Comparison of Undersampling, Oversampling, and SMOTE Methods for Dealing with Imbalanced Classification in Educational Data Mining

Educational data mining is capable of producing useful data-driven applications (e.g., early warning systems in schools or the prediction students’ academic achievement) based on predictive models. However, class imbalance problem educational datasets could hamper accuracy models as many these are designed assumption that predicted balanced. Although previous studies proposed several methods to deal with imbalanced problem, most them focused technical details how improve each technique, while only a few application aspect, especially for different ratios. In this study, we compared sampling techniques handle ratios (i.e., moderately extremely classifications) using High School Longitudinal Study 2009 dataset. For our comparison, used random oversampling (ROS), undersampling (RUS), and combination synthetic minority technique nominal continuous (SMOTE-NC) RUS hybrid resampling technique. We Random Forest classification algorithm evaluate results Our show seem work best. The implications suggestions future research discussed.