Musyimi Samuel Muthama, Dr. Kennedy Hadullo, Dr. Mvurya Mgala

Abstract

In the dynamic landscape of e-learning, precise prediction of student attrition holds paramount importance for educational institutions. This study addresses a significant research gap by introducing an innovative approach that seamlessly integrates Machine Learning, Naïve Bayes (Na), Distributed, Gradient Boosting, and Random Forest into a predictive machine learning model. Confronting the limitations of conventional models in capturing nuanced patterns associated with e-learning attrition, the methodology strategically leverages Simulated Annealing’s optimization capabilities and integrates ensemble learning for a distributed ensemble-based iterative classification model. Meticulous iterative adjustments of parameters and optimization of ensemble weights lead to the identification of the optimal model. Pioneering the application of Simulated Annealing (SA) contributes to heightened predictive accuracy, proposing a novel and optimal methodology to intricately address the challenges of attrition analysis and dropout prediction in e-learning environments. The research not only expands the methodological landscape but also offers a practical and optimal advancement for institutions striving to proactively identify and mitigate attrition.

KEYWORDS: – Attrition, attrition prediction, mitigation, E-learning, Machine-learning.