Data-Driven Student Support for Academic Success by Developing Student Skill Profiles

Abstract

In this paper, we attempt to provide a data-driven solution to the data-congested environment of attributes related to student success and contribute towards preventing the increased dropout rates at South African higher education institutions. One of the most significant discussions in higher education is student attrition in their first year of study. Student career guidance is an area that requires investigation in light of high attrition rates at university. Recent developments in data analytics, and the analysis of large data sets have enabled the production of powerful predictive models. This paper highlights how a predictive model can assist students, with an interest in Science to develop a skill profile required to be successful in their undergraduate Science programme. This is achieved by identifying the difference between the necessary skills required to be successful in a science programme (derived using data driven approaches) from the current learner's skill profile (derived from the learners' performance in assessments). The learners' skill profile is used to predict success in four Science streamlines. Based on the prediction results, we gauge the improvement in skills required to succeed in that programme. We provide the following contributions: (a) a trained classifier able to calculate the distribution over learners' success in Science streamlines focused around the notion of skill profiles; (b) a ranking of these skill profiles according to their information gain (entropy); and (c) an interactive program to calculate the posterior probability over these skill profiles given learner's pre-university observations. We argue that it is crucial that students gauge the focus areas of skill improvement prior to enrolling for their degree so that they can consider streams in Science degrees that are suited to their academic strengths.