Beyond service quality: a hybrid SEM-machine learning framework for human-centric passenger satisfaction in Addis Ababa's rail transit

Light Rail Transit (LRT) research has largely emphasized operational efficiency while overlooking passengers' psychological experiences. This study develops a human-centric framework integrating Environmental Sustainability Perception (ENV), Perceived Physical Health Benefits (PHB), and Mental and Emotional Well-being (MEWB) as drivers of passenger satisfaction. Using data from 1,273 Addis Ababa LRT passengers and a hybrid PLS-SEM -Machine Learning approach combining Random Forest, Deep Neural Networks, and SHAP explainability, the model explains 51% of the variance in passenger satisfaction. Results consistently identify MEWB as the strongest predictor (β = 0.421), substantially exceeding the effects of ENV and PHB across all analytical methods. However, MEWB and overall satisfaction recorded the lowest mean score (M = 2.40), revealing a critical well-being deficit. By integrating explanatory and predictive analytics, the study demonstrates that psychological well-being, rather than operational improvements alone, is essential for sustaining ridership and informing evidence-based transit planning in rapidly urbanizing African cities.