Search

Transport demand models are widely used to inform policy making and produce forecasts of future demand. A core output derived from demand models is the Value of Travel Time (VTT), which provides insights on the trade-offs that travellers are willing to make in terms of travel time and travel cost. VTT estimates are a critical input to cost-benefit analyses and feasibility assessments of potential projects and thus play a crucial role in transport planning and policy decisions. While much of the early work on VTT made use of revealed preference (RP) data, their use decreased due to growing concerns about reporting errors that may result in omitted observations and measurement errors in the model inputs. As a consequence, VTT measures have, for the last two decades, primarily been estimated using state-preference (SP) surveys. While SP methods can assess the individual trade-offs in a controlled manner, they are prone to behavioural incongruence. More recently, RP data from passively-collected data sources have raised the promise of accounting for some of the limitations of traditional RP surveys due to the minimal (or even no) active input from the respondent. The present study utilises such a dataset that combined a 2-week trip diary captured through smartphone GPS tracking with a household survey containing individual socio-demographic information. A mixed Logit model for mode choice was specified and the estimated parameters were then applied on the National Travel Survey to calculate the VTT estimates. Those estimates were further adjusted based on trip distances to get more representative national VTT values. This process resulted in estimates similar to the official UK guidelines used in transport appraisal that were obtained from SP data, where our results are not affected by concerns about response quality or survey artefacts. The findings hence strengthen the case for shifting towards passively generated RP data sources and are important for transport practitioners. [ABSTRACT FROM AUTHOR]

Departure time choice models quantify the relative impacts of the factors affecting travellers' departure time selection and help design targeted peak-spreading policies. The departure time preference of travellers is traditionally captured using parameters associated with different alternatives along three aspects – outbound, return, and duration. In reality, departure time decisions for outbound and return legs, and the corresponding activity durations, are interrelated in most cases. However, none of the previous departure time choice models has explicitly investigated the impact of this potential correlation on model outputs. To address this gap in the existing literature, we proposed a model structure with a novel polynomial functional form of alternative specific constants (ASCs) that captures this correlation in a joint (outbound and return) departure time choice model. A revealed preference (RP) dataset from Dhaka, Bangladesh, was used to model the joint departure time preferences of the car commuters. The proposed model was then compared with a state-of-the-art model that uses a trigonometric formulation of the ASCs. Results indicate that the proposed formulation yields more behaviourally realistic outputs compared to the trigonometric model by explicitly capturing the correlation between departure time and duration. While the specific outputs are applicable to car commuters residing in Dhaka, Bangladesh, the framework can be applied to better predict departure times and improve the formulations of the peak spreading policies in other contexts as well. [ABSTRACT FROM AUTHOR]

The rapid growth in passive mobility tracking technologies has led to departure time choice studies based on GPS data in recent years (e.g. Peer et al., 2013). GPS data however typically has limited sample sizes and is affected by technical issues like signal losses and battery depletion leading to gaps in the data. On the other hand, the rapid growth in mobile phone penetration rates has led to the emergence of alternative passive mobility datasets such as Global System for Mobile communication (GSM) data. GSM data covers much wider proportions of the population and can be used to infer departure time information. This motivates this research where we investigate the potential use of GSM data for modelling departure time choice. We describe practical approaches to extract relevant information from GSM data and propose a modelling framework that accounts for the fact that the desired departure times are unobserved. We assume that the preferred departure times vary randomly across the users and apply the mixed logit framework to jointly estimate the distribution parameters of the preferred departure times and the sensitivities to schedule delay. Comparison of the model results and time valuation metrics derived from the GSM data with similar metrics derived from the GPS data of a subset of the users reveals that the fewer time gaps in the GSM data lead to reliable model outputs. The proposed framework can be used for mobile phone and other passive data sources with unobserved preferred departure times. [ABSTRACT FROM AUTHOR]

In this research, we investigate the acceptability of three new and emerging smart mobility options and quantify the associated willingness-to-pay values in the context of Lisbon using a comprehensive stated preferences (SP) survey. The smart mobility options include shared taxi, one-way car rental, and a novel combination of park-and-ride and school bus facilities. While previous surveys on smart mobility options had investigated limited number of alternatives in isolation, the SP survey used in this research presents the smart mobility options alongside the existing options and their traditional variants like congestion pricing and improved public transport systems. Further, the choice of mode, departure time and occupancy are investigated in a multidimensional framework. This resulted a large choice set (with 9 modes, 5 departure times, and 2 occupancy levels leading to 135 alternatives in total) and required a novel survey design. The main survey administered over the internet and computer aided personal interviews included 2372 valid SP observations from 1248 respondents. Multi-dimensional mixed logit models were used to capture the complex correlations introduced due to the non-traditional survey design. Results indicate a significant preference of one-way car rental and shared taxi for non-commute trips. For commute trips, improved versions of traditional public transport modes are favoured over smart mobility options. These findings, as well as the novel data collection and modelling approach, are expected to provide important information to transportation planners and policy makers working to implement smart mobility options in Lisbon as well as in other cities. [ABSTRACT FROM AUTHOR]

In recent years, planners have started considering Machine Learning (ML) techniques as an alternative to discrete choice models (CM). ML techniques are primarily data-driven and typically achieve better prediction accuracy compared to CM. However, it is hypothesized that since the ML techniques do not have the strong grounding to economic theory as the CMs, they may not perform well in contexts that are radically different from the 'training' scenario. It is also hypothesized that the relative prediction performance may be affected by the metrics used for comparing the models. This research aims to test these two hypotheses empirically by modelling vehicle ownership choices using household survey data from Dhaka, Bangladesh collected in 2004, 2010 and 2019. The performances of CM (multinomial logit) and ML techniques (neural networks and gradient boosting trees) have been compared using log-likelihood and mean absolute percentage error of market shares. The results indicate that the multinomial logit model (MNL) with a piecewise linear transformation of the household income, has the best performance in terms of log-likelihood and mean absolute percentage error of market shares. This is followed by Neural Networks (NN) and Gradient Boosting Trees (GBT). The results thus provide empirical evidence that the ML techniques do not consistently outperform CM. Moreover, the difference in the performance of the models further increases if the prediction scenario is substantially different. This reinforces the hypothesis that CMs, with their behavioural underpinning, are better suited for long-term forecasting than data-driven ML approaches, especially if the population and network attributes are expected to change substantially. These findings will be useful for planners and policy makers in the selection of the appropriate tool for forecasting travel demand. [ABSTRACT FROM AUTHOR]