Modelling sustainable travel behaviour towards digital services: a machine learning-based framework

This study analyses the behavioural influences on the transport system shift of private vehicle users to digitally enabled, integrated multimodal transport systems in two major metropolitan areas in India. Rapid motorisation and increasing congestion make it necessary to understand the impact of digital service improvements and improved access in the first and last mile on the choice of mode and sustainable mobility transitions. A mixed revealed and stated-preference survey was performed. The perceived components were considered as predictors in a multinomial logit model and compared with several machine learning methods. Among all models, the deep learning model had the highest predictive accuracy (84%), indicating the model’s ability to capture complex, non-linear behavioural patterns. Factorial scenario simulations found that both digital service and multimodal access enhancements have much greater potential for mode shift than single interventions. The outcome highlights the critical role that the integrated strategy of seamless ticketing, real-time multimodal platforms, robust first- and last-mile connectivity and dynamic system management plays in reducing car dependence. This paper aligns with the broader paradigm of smart city development, and this argument supports the opportunity presented by technology-enabled mobility solutions to achieve sustainable urban transport transformation in the Indian context.