Promoting Sustainable Mobility: Understanding Commuter Mode Choices through Predictive Modeling

Abstract

Understanding commuters’ mode choice is crucial for promoting sustainable mobility and reducing car dependency. This study applies Multinomial Logit (MNL) and Neural Network (NN) models to survey data from employees in Rome, Italy, ensuring a fair comparison through identical preprocessing and evaluation metrics. Results show that while the NN model achieves slightly higher accuracy, statistical tests confirm the difference is not significant. Elasticity analysis in the MNL model examines key determinants influencing commuters’ decisions and provides interpretable insights into travel behavior. These findings demonstrate that the MNL model delivers strong predictive performance while maintaining greater interpretability. This reinforces the relevance of traditional econometric models in transportation research, particularly for policy applications where explainability is essential.