“…Even when using state-ofthe-art machine learning (ML) techniques, many models struggle to identify threshold effects, establish causality, and investigate interactions (Koushik et al, 2020). Utilizing the data-driven mechanisms of machine learning algorithms and corresponding interpretation techniques enables the analysis of feature importance, interaction effects, and partial dependence graphs of variables (Gao et al, 2021;Tu et al, 2021;Wei, 2022;Yang et al, 2021). Interpretable machine learning (IML) methods offer increased confidence in model outcomes, delivering higher prediction accuracy and superior performance in elucidating complex relationships.…”
The impact of the built environment and weather conditions on travel behavior has been widely studied. However, limited studies have focused on better understanding such effects in medium-sized cities with bus-oriented transit systems, particularly from a separate perspective of travelers’ origins and destinations. We took Weinan, China, as a representative of second-tier cities in developing countries that concentrate on bus-oriented development strategies. New evidence of feature importance and nonlinear effects of crucial factors were revealed by an interpretable machine learning-based approach combining XGBoost and Shapley Additive Explanation (SHAP) with multi-source data. Most key factors were critical at both origins and destinations, such as the density of residential and commercial facilities. However, several important factors, such as road density and boarding time, had strong imbalanced effects on travel behavior. These findings provide novel insights and empirical implications to support urban planning strategies in medium-sized cities.
“…Even when using state-ofthe-art machine learning (ML) techniques, many models struggle to identify threshold effects, establish causality, and investigate interactions (Koushik et al, 2020). Utilizing the data-driven mechanisms of machine learning algorithms and corresponding interpretation techniques enables the analysis of feature importance, interaction effects, and partial dependence graphs of variables (Gao et al, 2021;Tu et al, 2021;Wei, 2022;Yang et al, 2021). Interpretable machine learning (IML) methods offer increased confidence in model outcomes, delivering higher prediction accuracy and superior performance in elucidating complex relationships.…”
The impact of the built environment and weather conditions on travel behavior has been widely studied. However, limited studies have focused on better understanding such effects in medium-sized cities with bus-oriented transit systems, particularly from a separate perspective of travelers’ origins and destinations. We took Weinan, China, as a representative of second-tier cities in developing countries that concentrate on bus-oriented development strategies. New evidence of feature importance and nonlinear effects of crucial factors were revealed by an interpretable machine learning-based approach combining XGBoost and Shapley Additive Explanation (SHAP) with multi-source data. Most key factors were critical at both origins and destinations, such as the density of residential and commercial facilities. However, several important factors, such as road density and boarding time, had strong imbalanced effects on travel behavior. These findings provide novel insights and empirical implications to support urban planning strategies in medium-sized cities.
“…Below, when comparing ridership before and after the pandemic, we regard fiscal year 2019 as a pre-pandemic year and fiscal year 2020 as a pandemic year. Because precipitation and temperature affect bus ridership ( Hu and Chen, 2021 ; Wei, 2022 ; Kashfi et al, 2016 ), the daily temperature and precipitation data for Miyazaki City [ Table 2 (b)] were collected from the Japan Meteorological Agency. Fig.…”
“…The linkage disclosed new issues that have not been examined before. Theoretically, some new relationships related to human movements and the environment have been hypothesized and tested, for example, the interplay between travel behaviors and weather conditions or air-pollution degrees Wei, 2022;Xu et al, 2020; Methodologically, interdisciplinary dialogues have been introduced, participants of these dialogues including scholars from fields such as computer sciences, geography, computational social science, and urban studies.…”
Section: Embracing Uncertainty and Complexity Of Big Datamentioning
Considered a total social phenomenon, mobility is at the center of intricate social dynamics in cities and serves as a reading lens to understand the whole society. With the advent of big data, the potential for using mobility as a key social analyzer was unleashed in the past decade. The purpose of this research is to systematically review the evolution of big data's role in revealing social dimensions of urban mobility and discuss how they have contributed to various research domains from early 2010s to now. Six major research topics are detected from the selected online academic corpuses by conducting keywords-driven topic modeling techniques, reflecting diverse research interests in networked mobilities, human dynamics in spaces, event modeling, spatial underpinnings, travel behaviors and mobility patterns, and sociodemographic heterogeneity. The six topics reveal a comprehensive, research-interests, evolution pattern, and present current trends on using big data to uncover social dimensions of human mobility activities. Given these observations, we contend that big data has two contributions to revealing social dimensions of urban mobility: as an efficiency advancement and as an equity lens. Furthermore, the possible limitations and potential opportunities of big data applications in the existing scholarship are discussed. The review is intended to serve as a timely retrospective of societal-focused mobility studies, as well as a starting point for various stakeholders to collectively contribute to a desirable future in terms of mobility.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
