EVALUATING STUDENT OUTCOMES THROUGH PREDICTIVE MODELING: LESSONS LEARNED IN MALAYSIAN EDUCATION INSTITUTIONS

Abstract

Evaluating student outcomes through predictive modeling is pivotal for driving timely interventions and enhancing academic success in Malaysian education institutions. This systematic literature review (SLR), conducted following PRISMA guidelines, examined 106 studies published up to 2024 to assess the evolution and efficacy of predictive modeling techniques in this context. The review employed a structured search across Web of Science, Scopus, and Lens.org without publication year restrictions, ensuring comprehensive coverage of the literature. Studies were analyzed for their data collection strategies, and predictive methodologies, ranging from traditional cross-sectional analyses based on academic records to emerging longitudinal designs integrating multimodal data sources such as e-learning logs, surveys, and behavioral metrics. Findings reveal a dominant reliance on academic records, although recent trends indicate a gradual shift toward more sophisticated, multimodal approaches that enhance predictive accuracy. While conventional methods like decision trees and logistic regression remain prevalent, ensemble techniques, deep learning, and hybrid frameworks are increasingly adopted to mitigate challenges such as class imbalance and overfitting. The insights garnered underscore the potential of predictive modeling to inform early intervention strategies and policy decisions, highlighting the need for standardized, multi-institutional, and longitudinal research designs. By addressing these challenges, future research can better support data-driven initiatives that promote student retention, equity, and overall academic excellence in Malaysian educational institutions.