An interpretable hybrid activity-based and machine learning framework for modeling departure-time-dependent travel behavior in multimodal urban transport systems

Developing cities face increasing congestion and travel time uncertainty, yet conventional approaches often fail to capture the complex relationship between departure time, transport mode, and trip duration in multimodal urban systems. To address this limitation, this study proposes a hybrid Activity-Based Modeling–Machine Learning–Explainable Artificial Intelligence (ABM–ML–XAI) framework that integrates behaviorally realistic activity-travel simulation, predictive learning, and model interpretability within a unified framework. Unlike standalone ABM or ML approaches, the framework uses a behaviorally validated synthetic population derived from household travel survey data to generate a full synthetic spatiotemporal dataset capable of capturing nonlinear temporal–modal interactions in data-constrained urban environments. Using Bahir Dar as a case study, the full synthetic dataset generated in PTV VISUM was validated against observed travel statistics and used to train Random Forest (RF), XGBoost, and Multilayer Perceptron (MLP) models for trip duration prediction. Model evaluation incorporated R 2 , RMSE, MAE, residual analysis, mode-specific performance, and peak versus off-peak validation. The results reveal substantial travel time unreliability during congested periods, with a coefficient of variation of 0.91 and a planning time index of 3.30. RF achieved the best predictive performance (R 2 = 0.789, RMSE = 3.59 min, MAE = 1.87 min), outperforming XGBoost and MLP in capturing nonlinear travel behavior patterns. SHAP and LIME analyses revealed that transport mode was the most influential predictor of trip duration, followed by departure time, with strong nonlinear temporal effects observed during peak periods. The findings support evidence-based strategies for congestion mitigation, public transport reliability improvement, and sustainable urban mobility planning in rapidly urbanizing cities.