Short-Term Traffic States Forecasting Considering Spatial–Temporal Impact on an Urban Expressway

Abstract

Forecasting of short-term traffic states on expressways by adopting spatial–temporal models has gained increasing attention. Traffic data from neighboring sites were demonstrated to provide valuable information for predicting traffic states at sites of interest. However, when one considers the need to analyze the multivariate nature of traffic states over spatial dimensions, as well as of different models for various times of day, the interaction effects between spatial–temporal patterns require further investigation. This study addressed this issue on a segment of Shanghai North–South Expressway. Temporal characteristics of traffic volumes and speeds were analyzed by dividing the time of day into ordinal periods with relatively stable states. Then, spatial vector autoregressive (VAR) models were constructed at typical analysis periods for volume and speed forecasting by considering different combinations of upstream and downstream impacts. The results showed that the impact of downstream traffic conditions on upstream traffic cannot be neglected, especially in peak periods. For off-peak periods, traffic states at a location largely depended on upstream states, while downstream states appeared to have fewer effects. In such cases, models incorporating only upstream states were proved able to achieve sufficient accuracy. In addition, encouraging forecasting results were found when VAR models were compared with traditional methods (e.g., autoregressive integrated moving average and historical average), which failed to consider the spatial component of spatial–temporal patterns. All analyses helped to demonstrate the applicability of VAR models and to provide practical guidance for incorporating spatial–temporal dynamics into forecasting of expressway traffic states.