Predicting University Enrollment With Machine Learning: An Approach Using Random Forest And Extra Trees Classifier In The Peruvian Context

Abstract

In this study, we present the implementation of two predictive models based on Random Forest and Extra Trees Classifier for predicting university applicant enrollment. The main objective is to provide a support tool for selecting incoming students during the admission process. To this end, a dataset covering 10 academic semesters, from 2020 to 2024, was used and subjected to an exhaustive cleaning, preprocessing, and transformation process to ensure its quality and representativeness. The variables under study encompass academic, socioeconomic, and demographic dimensions, including preparation modality, career choice, type of school, parents' educational level, family income and expenses, employment status, among others. The dependent variable corresponds to the applicant's final enrollment status. The results obtained show that the Random Forest and Extra Trees Classifier models achieved high predictive performance, demonstrating a robust capacity to handle heterogeneous data and mitigate the risks of overfitting. Likewise, the levels of accuracy exceeded those of studies conducted in similar contexts, reinforcing their applicability in higher education. These findings support the potential of machine learning as an innovative strategy to strengthen equity and efficiency in applicant selection, providing scientific evidence of its usefulness in university higher education settings.