Predicting Academic Performance via Machine Learning Methods

Machine learning has been a heavily researched area in recent years, and many machine-learning methods for data analysis have been proposed in literature. The goal of this research was to explore various machine-learning methods for the purpose of predicting the future performance of Electrical Engineering majors based on their academic records from the common year in the College of Engineering. Machine-learning methods make predictions solely based on historical data, and no external biases are involved in the decision-making process. Therefore, such predictions can be much more objective than those offered through in-person meeting and “eyeball” tests. In our work, we used the final grades from ECEN 214 Electrical Circuit Theory as the primary indicator of future performance. Our research showed that both the Naïve Bayesian and Random Forest methods could lead to accurate predictions of the ECEN 214 final grade based on the student’s academic records from the common year. Our research also revealed the courses that have the most predictive power in the future performance of Electrical Engineering majors