Sequential Patterns for Spatio-Temporal Traffic Prediction

This paper proposes a method for predicting the traffic status of a city within time windows. The method takes advantage of space-partitioning, closed sequential pattern extraction, emerging pattern detection, and Markov chain modeling. From trajectories, we identify active regions in which moving objects mostly visit. The traffic status of each region is detected based on continuous tracking of closed sequential patterns evolution over time. Based on the proposed Markov model, the near-future status of traffic is then predicted. The traffic status is reported on maps and can be used to enhance future city transportation. The experiments on real-world data sets show that the proposed method provides promising results.