An Ensemble Modeling Approach to Enhance Grade Prediction in Academic Engineering Programming Courses

Abstract

Predicting the future academic grades of students can play a pivotal role in enhancing their performance in specific courses, consequently yielding a positive impact on their prospective academic, professional, and personal achievements, as well as on society at large. The field of programming is rapidly gaining prominence as an essential profession spanning multiple domains, marked by abundant opportunities and financial rewards. To cater to the diverse interests of students, the recommended curriculum structure for engineering programs in computing adeptly combines theoretical knowledge with practical programming skills. This approach ensures that students acquire a comprehensive understanding of programming courses, allowing them to choose the path that aligns best with their envisioned careers as programmers This research endeavors to introduce ensemble prediction techniques aimed at identifying students who exhibit the potential for advancement, or conversely, those who may not excel in four university-level programming courses. The outcomes of this study are presented alongside valuable performance assessment metrics for five ensemble methodologies, namely AdaBoost, Bagging, Random Forest, Stacking, and Voting. This evaluation employs a 10-fold cross-validation methodology and incorporates the Principal Component Analysis (PCA) for feature ranking. The results unequivocally demonstrate that both the Stacking and Random Forest ensemble approaches have attained the highest level of accuracy when applied to two distinct datasets.