Spatiotemporal Traffic Forecasting: Review and Proposed Directions
Recent working paper:
Ermagun, Alireza, and Levinson, D. (2016) Spatiotemporal Traffic Forecasting: Review and Proposed Directions (Working Paper).
This paper systematically reviews studies that forecast short-term traffic conditions using spatial dependence between links. We synthesize 130 extracted research papers from two perspectives: (1) methodological framework, and (2) approach for capturing and incorporating spatial information. From the methodology side, spatial information boosts the accuracy of prediction, particularly in congested traffic regimes and for longer horizons. There is a broad and longstanding agreement that non-parametric methods outperform the naive statistical methods such as historical average, real time profile, and exponential smoothing. However, to make an inexorable conclusion regarding the performance of neural network methods against STARIMA family models, more research is needed in this field. From the spatial dependency detection side, we believe that a large gulf exists between the realistic spatial dependence of traffic links on a real network and the studied networks. This systematic review highlights that the field is approaching its maturity, while it is still as crude as it is perplexing. It is perplexing in the conceptual methodology, and it is crude in the capture of spatial information.