Evaluation of Three Prediction Models for Identification of At-Risk First-Year Students

Abstract

Abstract
Education is a tool and means of achieving life goals that requires knowledge acquisition through diligence, practice and demonstration. Tertiary institutions adopt many techniques to educate students thus the success rates differ. Some students perform better when visual, textual or auditory based methods are used for teaching. Private tertiary institutions with limited student population are interested in retaining existing students’ and not expelling them. Additional services such as guidance, counselling and mentoring have been introduced to reduce failure  rates. These additional services are only useful if the weak students are identified early. In this work, data about first-year students in Computer Science was used to predict their academic performance and identify the students at-risk of failure in the first year. Naïve Bayes Algorithm, Decision Tree Algorithm and Support Vector Machine Algorithm were used to develop the predictive models. The results of the models demonstrated a prediction accuracy of 100%, a 0% classification error and a runtime of 505 milliseconds for the best model. Early identification of weak students will enable appropriate help to be activated for such students early in their academic life.