Quality data is a key to developing good transportation models. In this special issue on ‘Data Modelling in Transport,’ researchers tackle different issues in collecting, managing, and making use of different data types to enhance transportation modelling. In the first paper, Tam, Lam and Lo develop a discrete choice model with an attitudinal variable to better understand how passengers perceive service quality in ground access mode choices to the Hong Kong International Airport (HKIA). A structural equation model is adopted to construct a ‘satisfaction’ latent variable that describes the causal relationship between the satisfaction latent variable and the observable variables of the trip and of the passengers. This satisfaction latent variable is then incorporated into the discrete choice model as an additional explanatory variable to improve the significance of the choice model and to better explain the mode choice behaviour of departing air passengers. The model results indicate that inclusion of the satisfaction latent variable is not only important in improving the significance of the choice model, but also reveals that passengers do consider both observed service variables and unobserved attitudinal variable for their ground access to the HKIA.
In the second paper, Chang et al. provide a practical dynamic algorithm to forecast multi-interval bus travel time. The algorithm is developed based on the Nearest Neighbour Non-Parametric Regression (NN-NPR) using historical and current data collected by the Automatic Vehicle Location (AVL) technology. To show proof of concept, the algorithm has been applied to an intra-city bus route in Seoul, Korea. Preliminary results show that the algorithm performs well in both prediction accuracy and execution time, which are important for real-time applications.
The third paper by Lee et al. presents an interesting study of using the power prior to enhance road safety analysis. Specifically, the authors develop a full Bayesian approach with the power prior to estimate regression coefficients and effectiveness of road safety countermeasures. An estimation procedure based on the Metropolis–Hastings algorithm, one of the Markov Chain Monte Carlo methods, is suggested for Bayesian computation. A case study is conducted using two rural national roads in Korea. The results indicate that the full Bayesian approach performs favourably compared to the empirical Bayesian approach.
Automatic vehicle identification (AVI) has been considered as one of the emerging technologies used to collect real-time traffic information in a transportation network. In the fourth paper, Chen et al. provide three scenario-based multiobjective models to determine the locations of AVI readers under different travel demand patterns. The objectives are to consider the equipment and installation cost (e.g. number of AVI readers), the coverage of the AVI system (e.g. number of origin–destination (O–D) pairs), and the amount/quality of travel information obtained (e.g. number of AVI readings). To account for demand pattern uncertainty, each travel demand pattern is considered as a scenario in developing the scenario-based models. These three scenario-based multiobjective AVI reader location models are formulated as the expected value model, the minimum guarantee (maximin) model, and the minimum regret (minimax) model. A case study using the Irvine network in Orange County, California is provided to illustrate the applicability of the new models and the robustness of the AVI system designed by considering different travel demand patterns.
We would like to thank Prof. S.C. Wong, Editor-in-Chief of Transportmetrica, contributing authors and anonymous referees for helping to put together this special issue on ‘Data Modelling in Transport’. We hope that readers will enjoy reading this special issue.
Related Research Data
