AI-Based Early Detection of Student Dropout Risk Using Machine Learning on Academic and Behavioral Data

Abstract

Many students leave college or university before finishing their degree. This is called student dropout. Dropout is a serious problem because it affects the student's future and it also costs the university money and reputation. Traditional methods to find at-risk students, such as manual review by advisors, are slow and often find problems too late. This paper presents a simple framework that uses artificial intelligence (AI) and machine learning (ML) to find students who are at risk of dropping out at an early stage. The framework combines academic data, such as grades and attendance, with behavioral data, such as how often a student logs into the online learning system. Five machine learning models are compared: Logistic Regression, Random Forest, Support Vector Machine (SVM), K-Nearest Neighbors (KNN), and a small Artificial Neural Network (ANN). The models are trained and tested on a simulated dataset that reflects realistic patterns reported in prior dropout studies. Results show that attendance rate, number of failed courses, and weekly study hours are the strongest early warning signs. The Neural Network achieved the highest F1-score, while Logistic Regression achieved the highest precision. The paper explains the method in simple language so it can be understood by readers who are new to AI, and it discusses how universities can use this type of system responsibly to support students rather than to label or punish them.