A Prediction Model to Improve Student Placement at a South African Higher Education Institution

Abstract

There is a growing concern over the low pass rates of students in the Science Faculty at a South African Higher Education institution. The Admission Point Score (APS) used to place students into programs may appear to have good discretion in gauging student aptitude, but the reality is that between 2008 and 2015, about 50% of students who met the APS requirements for a Science program failed to meet the requirements to pass. This report attempts to build a recommendation engine that will advise students on their academic trajectory for a chosen program based on features suggested by the Tinto (1975) framework [1]. The results show that classification models from various archetypes of machine learning have good accuracy in predicting the final outcome of a new student. This research argues that a more complex view of student placement will improve the faculties success rates.