Accurate identification of economic hardship students: a data-driven approach

Abstract

Universities provide subsidies to students with economic hardship every year. However, the traditional method cannot accurately find those who really need help. Recently, students often use e-cards to consume in campus, and the consumption data from e-cards reflect the students' behaviours. In this paper, a data-driven method which combines statistical methods with machine learning classification algorithms (SM2L) is proposed to identify students with economic hardship by using consumption data. First, SM2L extracts seven features about meals and bath and excludes some abnormal consumption individuals according to consumption time and amount. Second, three classification algorithms are used to classify students and get different numbers of students with economic hardship by adjusting the parameters. Third, output the intersection of the result from phase 2 and the students who consume more in school. Experiments show that SM2L can accurately identify students with economic hardship.