Application of Logistic Regression Technique for Predicting Student Dropout
Berat Ujkani, Daniela Minkovska, Lyudmila Stoyanova
Abstract
Student dropout in higher education is a complex issue and as a process it includes many factors which may affect each other. This paper explores the use and application of a probabilistic supervised machine learning technique for predicting university student dropout to obtain insights on the students at risk and prevent them from dropping out the studies. Data from a public university in the Republic of Kosovo were obtained and examined. The dataset comprises instances of students' dropouts for the past six academic years along with their demographics, grades, enrollment details etc. Logistic Regression, as one of the most widely used Machine Learning and Artificial Intelligence algorithms, was used to build the model and produce the predictions. First, a statistical analysis was conducted and after the data preprocessing, logistic regression classifier was implemented. The results show that a high prediction accuracy was reached, with a percentage of 90%, and a F1 score of 0.85, indicating that the model is performing great and the predictions results are reliable.