Congestion pattern model for predicting short-term traffic decongestion times

Analysis of patterns in big traffic data can predict when road congestion events will dissipate. ITS organizations or drivers have interest in these pattern-based prediction about traffic congestion. In this paper, we propose methods for predicting traffic decongestion using congestion patterns. First, we propose a new method of representation using branched spatiotemporal chains. The method describes spatiotemporal changes in congestion for multiway branched roads. Second, we suggest a method for measuring similarities between patterns. This method can find the historical pattern that is most similar to the current congestion pattern. It then estimates the end time for the current congestion as that of the historical pattern. We performed experiments to compare the similarity of the estimated end times with real decongestion times for actual congestion events.