Performance Prediction for Undergraduate Degree Programs Using Machine Learning Techniques - A Preliminary Review

Abstract

Academic Performance prediction for undergraduate students is considered as one of the hot research areas since last couple of decades. An accurate and timely prediction of the student’s performance can directly influence the three participants; learner, instructor and the institution. This study presents a brief, preliminary review to explore existing literature from 2010 to 2022 in the context of performance prediction for Undergraduate Degree Programs (UDP). This review is organized according to Online and Traditional Education Systems (TES), and granularity level of performance output i.e., Degree program (Final CGPA), Next-semester, and the Course level grades. Aggregate analysis of the extracted data reveals that course level prediction is highly worked area deploying classification and regression techniques using data from academic domain. Existing empirical studies are mostly evaluated using accuracy, precision, recall and F1-measure and are validated with 10-fold cross validation. Contribution of this study is the novel categorical distribution of studies with respect to education system and granularity levels. Another important finding was the Success ratio of different Machine learning (ML) techniques used for these prediction studies. It is concluded that further research is required for TES to discover interdependent group of courses and Course Clusters for a certain degree program and then to develop prediction models for those course clusters.