A Hybrid Machine Learning–Land Use Interaction Model for Predicting Sustainable Transport Demand

Urban transport demand is intrinsically linked to land use patterns, socio-economic factors, and evolving mobility behaviors. Traditional transport planning models often fall short in capturing the complex, nonlinear relationships between land development and modal choices. This study proposes a novel hybrid modeling framework that integrates Machine Learning (ML) algorithms with Land Use–Transport Interaction (LUTI) models to predict sustainable transport demand in urban areas. The framework combines spatial land use data, accessibility indicators, socio-demographic variables, and historical transport usage patterns to train and validate predictive models such as Random Forest, XGBoost, and Multi-Layer Perceptrons. The ML component captures hidden patterns and nonlinearities, while the LUTI structure ensures spatial and planning-context relevance. A case study is conducted on a rapidly urbanizing Indian city to demonstrate the model’s performance in forecasting modal share shifts, evaluating transport sustainability metrics, and testing policy scenarios such as transit-oriented development (TOD) and congestion pricing. Results indicate that the hybrid model significantly improves prediction accuracy over conventional LUTI models and enables scenario-based policy testing for urban planners. This research contributes to the development of data-driven, adaptive tools for sustainable urban mobility planning in complex and dynamic urban systems. Keywords: Land Use–Transport Interaction (LUTI), Machine Learning, Sustainable Transport Demand, Urban Mobility Modeling, Random Forest, Transit-Oriented Development, Accessibility Index, Scenario-Based Planning, Spatial Data Analytics, Predictive Modeling.