Unlocking Student Success: Applying Machine Learning for Predicting Student Dropout in Higher Education

Student dropout remains a significant challenge for Higher Education Institutions (HEIs), affecting academic planning and student success. This study applies traditional machine learning and deep learning models to predict student dropout in an Ecuadorian HEI using the Cross-Industry Standard Process for Data Mining (CRISP-DM) methodology. A comprehensive analysis of demographic, academic, and economic factors was conducted to develop an effective predictive framework. The evaluated models include Logistic Regression, Support Vector Machine, Random Forest, XGBoost, Feedforward Neural Network, and TabNet. Various configurations were tested, including the application of Principal Component Analysis (PCA) for dimensionality reduction, and the Synthetic Minority Over-sampling Technique (SMOTE) to address class imbalance. Experimental results reveal that PCA and SMOTE are unnecessary. Among the models, Random Forest achieved the highest performance with a 96.62% accuracy, a ROC-AUC of 0.92, and an F1-score of 0.94. Feature importance analysis identified cumulative GP A and the number of semesters completed as the most influential factors for student dropout, followed by failed courses, high school grades, and entrance exam scores. This study emphasizes the importance of model interpretability, allowing HEIs to translate predictive insights into actionable strategies. By informing student retention policies and optimizing recruitment processes, this research contributes to data-driven decision-making in higher education.