Predicting the students with mental health risk by using Internet access logs

Abstract

Nowadays, the mental health problems of college students in our country are becoming more and more prominent. The mental health problems of college students not only hinder their healthy growth, but also affect the social and economic development of our country. Predicting students' mental health is an important field in educational data mining (EDM). However, it is very difficult to predict students' mental health because of many complex factors that affect the students' mental health, so currently there is little research on this field. As the Internet has almost become an essential part of students' life, the students' Internet use can reflect the students' psychological situation to some extent. Therefore, this study analyzes the online log of the freshmen students majored in computer in a university, and proposed an effective method to estimate the students' online time. Then, predict the students with mental health risk by using the students' online time on different types of Internet as features. The experimental results show that the proposed method is with high effectiveness and can predict about 50% of the students with mental health risk.