Prediction of main variables that lead to student dropout by using data mining techniques

Abstract

This research study aims to identify the main variables affecting student dropout. The behavior of student dropout (dependent variable) is predicted. There are 25 independent variables included that are grouped into five categories: personal, economic, social, family, and academic. These variables are sampled from a population of 410 students. The most influential variables are selected by using a multivariate statistical analysis. This generated a file structure that is analyzed using the Weka platform’s J48 algorithm. The results show that the most influential variables (effectiveness > 80%) for student dropout are: teacher pedagogy, frustration, the program’s importance, unmet expectations, program motivation, and procrastination. Variables such as academic average and age of admission appear not to be relevant. It is concluded that the obtained results provide valuable information for the deployment of university strategies that aim to reduce student dropout.