From urban transport model diversity to user preferences: A multilayer perceptron prediction

Abstract

The research addresses the complexity of urban mobility, highlighting the need to select the appropriate transport model under the user’s perception and under the sustainable development of modern cities. Achieving equitable, efficient, and environmentally responsible mobility systems necessitates collaboration among public and private sectors, complemented by active societal participation. Utilizing a dataset of 593 survey responses, a Multilayer Perceptron neural network was implemented to predict individual mobility preferences by integrating behavioral, demographic, and infrastructural determinants, including age, gender, occupation, car ownership, and Taxi/VTC usage frequency. Three primary mobility types were identified: public, shared, and private transport. The results indicate that car ownership and Taxi/VTC use are the most significant positive predictors of private mobility, whereas younger respondents exhibit a higher probability of adopting shared transport options. Methodologically, the application of neural network modeling enables the detection of nonlinear interactions and latent behavioral patterns often overlooked by conventional statistical approaches, thereby enhancing predictive precision and interpretability. These findings underscore the complex, multidimensional nature of mobility decision-making and highlight the utility of artificial intelligence techniques in advancing the analysis of travel behavior. The study’s implications extend to the formulation of inclusive, data-driven transport policies aimed at improving equity, accessibility, and sustainability in urban mobility systems, reinforcing the relevance of machine learning as a tool for evidence-based urban planning and policy development.