This paper proposes a model for the classification problem on imbalanced datasets, which uses a combination of the SMOTE model and the AdaBoost algorithm for the decision tree algorithm. We make a comparison between the proposed model and the decision tree algorithm using the Gini index and entropy on the collected datasets at Dong Hieu high school, Thai Hoa, Nghe An from 2014 to 2019. The research results can be used as a framework to develop applications supporting the early prediction of the ability of students’ dropout. Based on that results, the managers can analyze and come up with appropriate solutions in order to decrease the school dropout rate.