Accurate ride-hailing demand prediction is crucial for platform operational efficiency and transportation management. However, in emerging markets where historical trip data are entirely unavailable, conventional data-driven models as well as transfer learning approaches become infeasible, which creates a critical barrier for platform expansion and public-sector regulation. To bridge this gap, this study proposes a novel two-stage framework for zero-shot (i.e., target-data-free) cross-city ride-hailing demand prediction, in which total demand prediction is decoupled from its spatiotemporal allocation. In the first stage, a linear regression model estimates city-level aggregate demand using universally available macroeconomic indicators and built-environment features. In the second stage, a neural network learns transferable POI temporal activity patterns to allocate this demand across space and time, serving as a proxy for urban mobility rhythms without requiring target-city trip records. The framework is validated on a dataset of 5 million ride-hailing trips from 10 cities under a strict zero-shot experimental design. Results show that the method achieves superior spatial transferability and significantly outperforms baseline models, particularly when transferring knowledge from large, data-rich cities to smaller, data-scarce ones. The predicted demand accurately captures key spatiotemporal dynamics—such as morning and evening peaks—and the learned POI activity profiles are highly interpretable, aligning with known urban activity schedules. This study contributes significantly to both research and practice. Methodologically, it presents a robust and transferable solution for demand forecasting under complete data scarcity. Experimental results demonstrate that the proposed method reduces RMSE by up to 64.55% and MAPE by up to 66.31% compared to baseline models, particularly when transferring knowledge from data-rich metropolises to data-scarce emerging cities. For platform operators, it provides an actionable tool for strategic market entry, fleet sizing, and risk assessment. For transportation planners and policymakers, it enables proactive evaluation of traffic impacts, coordination with public transit, and evidence-based regulatory design in emerging mobility markets.
Influenced by sociocultural norms and ongoing safety shortcomings in urban transportation systems, women frequently perceive a higher personal risk during routine travel. Building on the concept of the “pink tax” in consumer markets, this study introduces the notion of a transportation pink tax as a gendered safety-compensation burden, the additional expected monetary cost that women incur when safety concerns prompt compensatory behaviors, such as choosing more expensive travel modes. Using a stated preference (SP) experiment conducted in Shanghai (N = 785; 391 females and 394 males), we examine mode choice among metro, ride-hailing, and autonomous robotaxi services. Both multinomial and mixed logit models are employed to estimate gender-differentiated preferences, derive willingness-to-pay (WTP) for safety-related attributes, and account for unobserved heterogeneity. The results indicate that women place significantly higher marginal utility on safety-enhancing features, including female ride-hailing drivers, advanced in-vehicle security systems, and improved lighting in robotaxi access segments, reflecting a greater safety-compensation premium in their utility functions. For example, women’s WTP for advanced security measures in robotaxi services exceeds that of men by 3.90, 26.05, and 27.86 CNY for short-, medium-, and long-distance trips, respectively. Importantly, we show that this marginal safety premium translates into a tangible cost burden when women shift toward higher-cost modes under unsafe conditions. Policy scenario simulations indicate that enhancing metro safety, through measures such as shorter walking distances, upgraded in-carriage CCTV, additional security staff patrols, and improved lighting, could raise women’s metro usage and reduce their monthly travel expenditures by approximately 26–121 CNY, thereby alleviating the gendered safety-compensation burden. These results suggest that transportation gender inequities arise not from inherent preference differences, but from structural safety externalities that disproportionately impact women. Implementing safety-by-design principles in public transit and emerging autonomous mobility services can therefore lower gendered cost burdens and foster more equitable urban mobility.
Modeling non-work activity destination choices is essential for urban transportation planning, as these activities involve greater flexibility and more complex behavioral decision-making than mandatory activities. However, existing studies often treat all facilities within the same activity category as homogeneous and have rarely examined how facility-grade heterogeneity shapes individuals’ destination preferences. To address this gap, this study develops a hierarchical location choice modeling framework that explicitly incorporates facility-grade distinctions into the destination choice process. Four common non-work activity categories—shopping, dining-out, recreation, and medical care—are analyzed, with urban facilities classified into high-grade and low-grade types. Destination choice sets are generated using fuzzy logic inference and stratified importance sampling that account for both travel distance and facility grade. Conditional logit (CL) and nested logit (NL) models are then estimated to capture alternative-specific utilities and within-grade correlations. Results show that (i) destination choices exhibit distinct decision-making logic across facility grades, with high-grade facilities usually planned for quality-seeking purposes and low-grade facilities primarily serving convenience-oriented needs; (ii) facility-grade heterogeneity significantly shapes destination choice behavior, with preferences varying across activity purposes and population groups; (iii) the nested logit results reveal correlated destination choices within facility-grade categories, highlighting the importance of accounting for within-grade substitution patterns when modeling non-work destination choices; and (iv) the effects of accessibility improvements vary across activity types, with stronger impacts observed for shopping and recreation and comparatively limited impacts for dining-out and medical-care. These findings highlight the critical role of facility-grade heterogeneity in shaping non-work travel behavior and offer new insights for demand analysis and facility planning.
Declining revenue from traditional road funding sources, rising infrastructure costs, and the transition to electric vehicles have increased the urgency of road user charging reform as a key demand management strategy. While such schemes can improve transport system efficiency by pricing congestion and other externalities, their implementation requires careful balancing with concerns around fairness and affordability. Public acceptability therefore remains a critical constraint on policy adoption. This paper examines which policy features most strongly influence support for road user charging and how these preferences can inform the design of policies that advance efficiency, fairness, and affordability objectives. A best worst scaling approach is used to elicit the relative importance of policy features, with choices modelled using a hybrid choice framework that captures both observed preferences and underlying attitudes. Three distinct behavioural classes are identified, reflecting differing priorities related to efficiency, fairness and consistency, and broader public benefit. Across these groups, governance and institutional arrangements are central to perceived legitimacy. Features such as public ownership, not for profit operation, independent investment decision making, and transparent revenue use are strongly preferred, while more complex pricing mechanisms are viewed less favourably. The findings highlight the importance of trust, fairness, and affordability in supporting effective and acceptable reform.
Digital ticketing is increasingly central to smart urban mobility, yet its operational benefits depend on passenger adoption and institutional readiness. This study examines the impact of digital ticketing on passenger-flow efficiency and user acceptance at BTS Khu Khot Station, a major suburban terminal in Bangkok. A sequential mixed-methods design was adopted. First, Arena-based discrete-event simulation assessed three digital adoption scenarios involving Rabbit cards and ticket vending machines. Second, passenger survey data (n = 402) and expert responses (n = 12) were analysed using constructs informed by TAM and UTAUT2. Simulation results show that increasing non-manual ticketing adoption from 45% to 80% reduced average system time from 4.55 to 2.86 min and queue time from 2.12 to 1.05 min, while throughput increased from 2.51 to 3.87 entities per minute. The findings highlight digital ticketing as both an operational intervention and a user-centred implementation challenge.
Feebates applied to new vehicles are widely used to reduce emissions from the automobile sector, yet most vehicles are traded in second-hand markets. This paper studies how interactions between new and used car markets affect the distributional incidence of a car feebate. Using comprehensive French administrative data on vehicle transactions over the period 2008–2019 and detailed information on used car prices, we develop an accounting framework linking the feebate spillover to vehicle depreciation.We find that over half of the initial feebate is redistributed through used car markets within five years, with important consequences for incidence. Corporate buyers, who face the largest direct tax on new cars, recover approximately two-thirds of their burden through resale and end up close to neutral overall. Among private households, the burden shifts further towards high-income groups, who purchase both more new and more recent used cars. Yet the feebate remains progressive: the net burden as a share of income increases with household income, whether accounting for used car markets or not. These results suggest that the distributional effects of vehicle regulations cannot be assessed on the basis of new car markets alone, and that corporate buyers play a key role in transmitting feebate costs across the car lifecycle.
Carpooling is widely regarded as an effective approach to mitigating urban traffic congestion and environmental pollution. However, adoption remains limited, and rigorous causal evidence on carpooling regulatory policies (CRP) is scarce. Using high-resolution order data from 28 cities in three major Chinese urban agglomerations in 2017, this study exploits staggered city-level CRP implementation as a quasi-natural experiment to estimate its effects on carpooling activity. We develop a spatiotemporal feature-based classification algorithm (SFBCA) to distinguish LF-HSC drivers, characterized by lower carpooling frequency or higher spatiotemporal concentration, from HF-LSC drivers, characterized by higher carpooling frequency and lower spatiotemporal concentration. We estimate average effects using a two-way fixed-effects staggered difference-in-differences model and dynamic effects using heterogeneity-robust event-study estimators, supplemented by robustness checks. Under the monthly specification, SFBCA achieves consistency rates of 93.8% for LF-HSC drivers and 73.0% for HF-LSC drivers, yielding a harmonic consistency of 82.1%. CRP reduce total carpooling orders by 10.2% and HF-LSC drivers’ orders by 25.0%, while having no statistically significant average effect on LF-HSC drivers’ orders. These findings suggest that CRP primarily constrain more intensive and potentially quasi-commercial carpooling activity without producing a detectable average reduction in lower-frequency or more spatiotemporally concentrated carpooling. Heterogeneity analyses reveal variation across urban agglomerations, policy designs, and market environments, with statistically detectable effects concentrated in cities with standalone and detailed regulations. Overall, the findings provide implications for governments and transportation network companies (TNCs) to refine carpooling regulation through targeted and context-sensitive implementation.
This paper proposes a hybrid approach combining the Gravity Model (GM) with machine learning (ML) techniques to improve predictions of Malaysia's international freight exports and imports. The study evaluates GM and three ML models; Random Forest (RF), Artificial Neural Network (ANN), and Support Vector Machine (SVM), with RF emerging as the most robust performer based on RMSE, MSE, and R-squared metrics. Four hybrid approaches were developed by combining GM and RF predictions with residual modeling. Results demonstrate that hybrid models consistently outperform standalone approaches. Notably, Hybrid Approach 2, which uses GM predictions adjusted by RF-modelled residuals with RF features, delivered superior performance, achieving RMSE values of 0.226 for exports and 0.162 for imports. This confirms the hybrid model's enhanced capability to capture complex trade patterns, offering a valuable forecasting tool for Malaysia's international freight transport planning.
This study aims to identify and prioritize green urban freight transport (UFT) initiatives by evaluating the barriers associated with their implementation. A hybrid multi-criteria decision-making (MCDM) model combining Analytic Hierarchy Process (AHP) and Fuzzy, FAHP was employed. Thirty-two initiatives were categorized into six groups, and nine barriers were identified through a comprehensive literature review. AHP was utilized to collect expert opinions, while FAHP ranked the initiatives and barriers in the context of Hanoi, Vietnam, an emerging economy with increasing concerns about air pollution from transport activities. The findings highlight five highly feasible initiatives: raising awareness of green transport, practicing eco-driving, extending collection point operating hours, promoting electric delivery motorcycles/bicycles, and developing e-commerce pickup networks. In contrast, infrastructure-related initiatives face significant challenges due to high resource demands. From an academic perspective, this study reinforces the applicability of MCDM approaches in addressing the complexities of urban transport systems, while contributing to the understanding of challenges and opportunities for sustainable urban freight systems in emerging economies. The findings also provide a valuable insights for urban transport managers, policymakers, and freight operators.
Shared micromobility has emerged as a crucial component of urban transportation, yet high operational costs remain a significant bottleneck. User incentives, primarily offered as fare discounts, are widely adopted to mitigate these issues; however, their effectiveness relies heavily on a thorough understanding of user choice behaviors. This study investigates user responses to incentive mechanisms in shared micromobility using a novel large language model (LLM)-based behavioral simulation framework. Profile-based user digital twins are developed to emulate responses under incentive scenarios at scale. Two types of incentives are analyzed: station-based incentives (spatial relocation) and vehicle-based incentives (battery adjustment). The fidelity of these digital twins is evaluated across dimensions of stability (consistency testing), rationality (rationale analysis), and validity (empirical benchmarking), suggesting their ability to capture key stated preference patterns under controlled scenarios. Findings highlight user heterogeneity, with younger, lower-income, and non-commuting individuals showing greater acceptance than older, higher-income, and commuting users. Past interactions influence incentive effectiveness, as prior acceptance promotes engagement and negative experiences deter it, with both effects initially strengthening before diminishing. However, exposure frequency has no significant impact on incentive acceptance. Vehicle-based incentives can serve as an efficient baseline mechanism for routine operations with lower adoption barriers. In contrast, station-based incentives encounter greater initial resistance, but are more responsive to financial rewards and suited for targeted rebalancing. The work offers practical insights for incentive design in shared micromobility and demonstrates the potential of LLMs in behavioral simulation for transportation.
Recent advances in autonomous driving have accelerated the deployment of robotaxi services, positioning them as a key component of future urban mobility. However, existing research has largely emphasized pre-adoption intentions, offering limited insight into users’ post-experience evaluations and continued usage behavior in fully driverless contexts. To address this gap, this study examines how perceived value and perceived risk, conceptualised as higher-order constructs, are jointly associated with continued usage intention toward robotaxi services in China. Guided by the cognitive–affective–conative (CAC) framework, perceived value and perceived risk are conceptualized as multidimensional cognitive evaluations formed through actual usage experiences. Based on survey data collected from 409 users with real-world robotaxi experience in Wuhan, China, the study employs structural equation modeling to investigate how these cognitive evaluations shape affective responses, including positive emotion, negative emotion, and satisfaction and, in turn, continued usage intention. The results show that perceived value is strongly associated with higher post-experience positive emotion and satisfaction, which in turn are positively related to continued usage intention, with satisfaction emerging as the strongest direct predictor. In contrast, perceived risk is primarily associated with higher post-experience negative emotion, while its associations with satisfaction and continued usage intention remain weak. This suggests that positive affective and evaluative mechanisms play a more dominant role in sustaining usage than cost-related concerns. By extending the CAC framework to post-experience robotaxi usage, this study deepens the understanding of continued usage behavior in autonomous mobility and offers practical implications for enhancing user acceptance and supporting the large-scale deployment of robotaxi services in China.
To support bicycling, cities have focused on providing infrastructure to ensure safe movement. Much less research assesses the provision of adequate bicycle parking, especially residential parking. Because of their higher weight and costs, e-bicycles especially require safe, sheltered and accessible parking.The paper describes the features of home-based bicycle parking for Swiss individuals with and without e-bicycles and assesses variations and discrepancies in residential bicycle parking availability and e-bicycle ownership across land use and household characteristics.The 2021 Swiss Mobility and Transport Microcensus (MTMC) provides a representative national sample of data on e-bicycle ownership and residential bicycle parking (n = 10,154). Variations in logit-based estimated probabilities of e-bicycle ownership and quality parking access across predictor variables are assessed.The probability of having accessible sheltered and secured bicycle parking in a locked room, the golden standard of safety, protection from the elements and ease of use, is low and below that of e-bicycle ownership probabilities, regardless of socio-demographics and location features. Across group characteristics, specific mismatches between parking access and e-bicycle ownership are identified. The largest mismatches are found for homeowners and residents of areas poorly served by public transit.Current housing stock is ill-equipped for parking of e-bicycles, especially in areas where regular cycling is historically low. While existing standards can help shape new housing stock, retrofitting existing residential buildings is an important policy challenge. On street veloboxes may be more feasible. The variety of observed bicycle parking forms also points to the limits of current survey instruments. Improvements to survey instruments are proposed.
This study introduces a novel design-of-designs survey framework tailored to model decision-making processes involving the presence of alternative choice heuristics. Often based solely on a utility maximization framework, traditional choice models fail to account for the diverse heuristics that individuals may employ when faced with varying task complexity and choice contexts. The design-of-designs framework allows capturing potential alternative heuristics by systematically varying task complexity along meta-dimensions such as the number of alternatives, attributes, and choice contexts in different transport-related choices (of route, holiday destination, and residential location).The survey framework incorporates advanced features, including dynamic presentation schemes, perceptual indicators, and cognitive effort metrics, to elucidate how survey complexity may shape heuristic adoption. In particular, residential location alternatives are displayed on interactive maps, enabling respondents to filter alternatives based on selected attribute values and providing insights into selective information processing and the use of choice paths. Additionally, stated non-attendance indicators and response times are registered to evaluate cognitive load and attribute importance. Finally, the survey design also includes the possibility of considering latent constructs, such as engagement, thinking styles and trust in institutions, which have proven important in modelling with alternative heuristics.The paper highlights the importance of integrating non-compensatory behaviour into discrete choice models to enhance behavioural realism by capturing the dynamics of heuristic-driven decision-making. Finally, the paper discusses how the different survey elements can support future research on information processing and heuristic use in stated choice surveys.
Understanding how built environment characteristics influence electric vehicles (EV) idling emissions is crucial for designing low-carbon transportation systems, yet this stationary emission source remains largely overlooked. This study fills this gap by integrating high-resolution GPS trajectory data, grid carbon intensity, and explainable machine learning to quantify the nonlinear impacts of the built environment on EV idling emissions. We developed a LightGBM-SHAP framework that achieves robust predictive performance (R2 = 0.889) and systematically identified nonlinear thresholds. Three key findings emerged. First, across the seven observation days, lower daily mean temperatures were generally associated with higher estimated daily CO2 emissions under the adopted temperature-dependent emission-accounting framework. Second, road network length, residential land, transportation facilities, and population density collectively accounted for nearly 80% of total SHAP importance, with multiple built environment features exhibiting nonlinear threshold effects. Third, feature significance exhibited spatiotemporal heterogeneity, with identical built environment configurations exerting markedly different impacts across varying time periods and spatial scales. These findings should be interpreted as predictive associations rather than causal planning standards, offering diagnostic evidence for future policy evaluation.
