An Integrated Predictive-Analytical Modeling Framework for E-Learning Quality Assessment: Evidence from the TBSA Program
Keywords:
E-Learning Quality, Machine Learning, SHAP Analysis, Fuzzy TOPSIS, Hybrid Methodology, Learner Satisfaction, Educational AnalyticsAbstract
The rapid expansion of e-learning necessitates robust frameworks to evaluate and enhance learner experience. This study proposes an innovative hybrid methodology that integrates machine learning (ML) with fuzzy multi-criteria decision-making (Fuzzy MCDM) to assess perceived e-learning quality. Using data from the TBSA (Training for Business Start-up Advisors) program, we first employ Gradient Boosting with SHAP (SHapley Additive exPlanations) analysis to objectively determine feature importance from learner responses. These data-driven weights are then integrated into a Fuzzy TOPSIS model to manage inherent uncertainties in educational assessments and produce robust quality dimension rankings.
Our results reveal that Responsiveness/Ease of Use (dimension weight: 0.5714) is the most critical dimension, followed by Tangibility (0.2020), Assurance (0.1592), and Security/Reliability (0.0674). The hybrid framework demonstrates substantial explanatory power (R = 0.5156) while providing interpretable, action- able insights for e-learning optimization. This approach offers educational institutions a scientifically-grounded methodology for prioritizing quality improvements and resource allocation.
