Dr. Qiang Meng is currently a Professor in the Department of Civil and Environmental Engineering (CEE) at National University of Singapore (NUS), the co-director of LTA-NUS Transportation Centre as well as the director of Centre for Transportation Research (CTR) of CEE. His research mainly focuses on urban mobility modeling and optimization, shipping and intermodal freight transportation analysis, and quantitative risk assessment of transport operations. He has published more than 210 articles in the leading transportation and logistics journals, with the H-index rate of 64 and the total citations of 12,164 in Google Scholar.
Dr. Meng is currently the Co-Editor-in-Chief of Transportation Research Part E and Associate Editor of Transportation Research Part B. He has clinched a number of research awards and prizes, including The 2020 TSL (Transportation Science & Logistics Society of INFORMS) SIG Best Paper Award in Freight Transportation and Logistics in 2020, the OCDI Takeuchi Yoshio Best Paper Award in the Field of Logistics in the 13th EASTS International Conference in 2019, Engineer Research Award of Faculty of Engineering at NUS in 2018, Outstanding Alumni Award of the Department of Civil and Environmental Engineering at The Hong Kong University of Science and Technology in 2016, Dean’s Chair in Faculty of Engineering at NUS in 2015, the 13th World Conference on Transportation Research (WCTR) Society Prize for the best paper in 2013.
Traffic state prediction is gaining more and more attention worldwide in recent years. Many published studies have investigated the various types of models used in traffic state predictions. However, there lacks a systematic and critical review of the key components of the proposed models. Therefore, this study reviews the machine learning-based traffic state prediction models. Three types of machine learning models are retrieved and reviewed, namely, the statistic-based models, the instance-based models, and the neural network-based models. A framework regarding the key components of the traffic state prediction models is proposed. This framework comprises (i) road network representation for the geospatial information of the road network, (ii) essential learning features that extract the features/characteristics behind data, for instance, spatial/temporal data correlations, and (iii) model structure, which focuses on designing the model layout for the features extracted to be fully learned by the model. The existing studies are thoroughly examined according to these components. Open challenges are discussed and future research directions are provided.