Enhancing Global Student Success Through Data-Driven Session Design in Online Education

Abstract

Effective online learning sessions require designing sessions that address learners'engagement and achievement across diversegroups. The duration and frequency of the session affect user satisfaction and quiz results, but it is difficult to optimize both simultaneously.This paper presents a combined optimization model that uses stepwise regression, NSGA-II, and gray relational analysis (GRA) to optimize the design of a session, leveraginga publicly available e-learning dataset (more than 2,500 anonymized records). Directional relationships between session parameters and outcomes were quantified using regression models,and Pareto-optimal solutions were identified using NSGA-II, which were further assessed under three teaching-priorityscenariosusing GRA. The results show that a 60-minute weekly session is theoptimal balance between the more satisfaction-focused designs, and that allotting 113 minutes across eight sessions is optimal for quizperformance. The explanations of the regression models (R2 = 0.20 for satisfaction and R2 = 0.11 for quiz scores) are modest, suggesting that the results should be viewed as guidance for decision-making rather thanprescriptions. Despitethese shortcomings, the framework emphasizes trade-offs between the timingand frequency of online learning and offers a data-driven,systematic approachtooptimizing online learning. Thisresearch contributesto evidence-based instructional design and provides practitioners with actionable insights to enhance international online learning.