Predicting Student Success in Higher Education with Social, Economic, and Education Variables Using Machine Learning Techniques

Student success is a topic of extensive literature, given the importance it represents for educational institutions to be able to detect and act preventively on students who are at risk of not approving or withdrawing from their studies. Precisely, in higher education, student success can be measured by the successful graduation of students because if a university student fails to approve all the courses in their curriculum, they will not be able to access their degree. Higher education establishments maintain academic data repositories encompassing student records, grades, course enrollment, and scholarship allocations. Institutions leveraging Learning Management Systems can integrate learning process data to predict student success through Learning Analytics methodologies. Past studies have correlated collegiate success with academic prowess evaluated via aptitude assessments or prior high school performance; our study extends this paradigm by incorporating socioeconomic variables. We contend that a cohort of students, despite possessing adequate academic aptitude, faces challenges leading to study withdrawal or non-completion. Our investigation focuses on the Salesian Technical University (UPS), a nonprofit private institution that employs a socioeconomic quintile system akin to the World Bank Group's classification, allotting scholarships to students from disadvantaged backgrounds. Our research uses correlation analyses among diverse academic, social, and economic variables to discern the most influential factors in predicting academic success. Leveraging these correlations, predictive models employ machine learning techniques. The culmination of our study reveals the identification of student cohorts at academic risk, often overlooked by traditional analytic approaches reliant solely on academic data.