Views: 122 Downloads: 27

Abstract

This study presents a solution for predicting university dropout rates, leveraging advanced digital technologies and the Random Forest algorithm. The model was developed using key academic variables, such as year of enrollment, program of study, semester attended, and academic performance, represented by the grade point average (GPA). A dropout threshold was established for students whose GPA fell below 11. The dataset was partitioned into 70% for training and 30% for testing, yielding an overall accuracy of 85.9%. Feature importance analysis identified semester and year of enrollment as the most influential factors in predicting dropout. While the model demonstrated a 91% accuracy in identifying students unlikely to drop out, its predictive capacity for students at risk of dropping out was limited to 52%. This approach constitutes a significant advancement in the implementation of digital technologies in education, enabling proactive strategies to improve student retention through data-driven predictive interventions‎.

Keywords: University dropout, Prediction, Random forest, Academic performance, Retention

Citation Formats: