An Investigation into Student Performance Prediction using Regularized Logistic Regression

The problem of university dropout poses a significant challenge to education systems worldwide, affecting administrators, teachers, and students. Early identification and intervention strategies are crucial for addressing this issue. In addition, advances in machine learning have paved the way for more accurate predictions of student performance. This paper investigates the use of regularization techniques, specifically Lasso (L1) and Ridge (L2) regularization, within logistic regression models to improve the accuracy of performance prediction. This research’s dataset was obtained from the Binus Online Learning platform at Bina Nusantara University, with a focus on the Information System study program between 2020 and 2021. The results reveal that logistic regression with regularization achieves a high level of accuracy, recall, and precision in predicting student performance. The findings contribute to the development of an early warning system to identify at-risk students, enabling timely intervention and support.