Intelligent Prediction of EFL Student Performance in Higher Education Using Ensemble Machine Learning Techniques

Abstract

Digital technologies have rapidly changed teaching and learning, forcing educators to use novel methods to improve student engagement and academic performance. EFL Student Performance in Higher Education has shown that gamification may motivate and engage students in language learning on computer and mobile platforms. Due to the richness and diversity of educational data, predicting student success is difficult. This article provides an intelligent prediction framework using the ResXGR hybrid model to address this issue. Preprocessing the educational dataset with Data Cleaning and Z-Score Normalisation improves data quality and consistency. HCFS is then used to extract the most important student performance features. The ResXGR architecture uses SE attention mechanisms to capture high-dimensional academic characteristics and the Jaya optimisation algorithm to optimise model hyperparameters for predictive accuracy and computing efficiency. Experimental results show that the proposed model outperforms conventional ML and contemporary prediction algorithms with 96.28% prediction accuracy, faster convergence, and lower processing overhead. These findings demonstrate that the proposed framework is scalable, robust, and efficient for predicting EFL Student Performance in HE, enabling timely academic interventions, informed educational decision-making, and improved student learning outcomes in technology-enhanced EFL environments.