The academic success of students is the top-priority objective of all institutions across the globe because it is a widely accepted indication of their overall reputation and ranking in the educational field. Institutions can benefit significantly from academic performance prediction models, which can assist them in providing excellent resources and learning experiences to students, thereby ensuring their retention in the academic arena. As such, predicting student academic performance using educational datasets has become one of the most essential components of the technology-assisted educational environments of the present times. This study proposes a robust prediction model using the Extreme Learning Machine (ELM) and Random Forest (RF) complementary strengths. Using ELM facilitates quick learning speed and underlines the ability to efficiently handle large-scale data to map input data features. The RF classification technique, noted for its accuracy and capacity to handle complicated feature interactions, classifies the mapped features. The E-RF hybrid model involving ELM’s feature extraction and RF’s classification capabilities yields improved prediction accuracy and computational efficiency, as is evident from the experimental results.