Dropout early warning systems for high school students using machine learning

Abstract

Students' dropouts are a serious problem for students, society, and policy makers. Predictive modeling using machine learning has a great potential in developing early warning systems to identify students at risk of dropping out in advance and help them. In this study, we use the random forests in machine learning to predict students at risk of dropping out. The data used in this study are the samples of 165,715 high school students from the 2014 National Education Information System (NEIS), which is a national system for educational administration information connected through the Internet with around 12,000 elementary and secondary schools, 17 city/provincial offices of education, and the Ministry of Education in Korea. Our predictive model showed an excellent performance in predicting students' dropouts in terms of various performance metrics for binary classification. The results of our study demonstrate the benefit of using machine learning with students' big data in education. We briefly overview machine learning in general and the random forests model and present the various performance metrics to evaluate our predictive model.