Machine learning techniques for evaluating the nonlinear link between built-environment characteristics and travel behaviors: A systematic review

Abstract

For years, linearity assumptions have been used to assess the relationship between the built environment and travel behavior. However, recent investigations have revealed that these interactions are often nonlinear. As a result, machine learning (ML) techniques are being used to model these nonlinear interactions instead of traditional statistical methods. This has the potential to improve research into the relationship between the built environment and travel behavior. However, while there have been rigorous reviews of traditional methods for assessing linear impacts, the technical aspects of applying ML methods to evaluate nonlinear connections have not been thoroughly reviewed. To address this gap, this study examines the technical aspects of applying ML methods to analyze nonlinear correlations between the built environment and travel behavior. The review identifies state-of-the-art strategies for modeling the relationship between the built environment and travel behavior using ML techniques and highlights methodological shortcomings in previous research. The review is based on a comprehensive search of three main online publishing sources, which yielded 41 distinct papers. The review extracted 11 qualities encompassing five research questions: application domains, ML approaches, datasets, performance assessment, and hyper-parameter optimization. Additionally, the study identified ten technical issues that may impact the effectiveness of the ML models. Of these, three were identified as technical matters of concern that may introduce bias in the evaluation of model effectiveness. Seven were identified as potential enhancements that may improve the model's effectiveness. Finally, four more general issues were identified as potential areas for future research.