Early detection of the academic performance of university students from first year through discriminant analysis

Abstract

This work aims to identify students who would only pass at most two out of five enrolled courses from semester 2016-2 of the General Studies Program of Universidad de Lima. The study is based on predictive models constructed with data collected on semester 2016-1 through discriminant analysis. The student’s population was divided in three domains of study. Then, independent predictive models for academic performance were constructed using Fisher’s classification functions which were evaluated by performance indicators and the Receiver Operating Characteristic curve (ROC).