Medium-term prediction of urban traffic states using probability tree

Prediction of the future traffic states of the overall city arterial networks provides the process of urban traffic evolution for researchers, which can serve as reference information for the city traffic authorities. However, the complexity and heterogeneity of urban traffic system and the big data challenge have proven to be substantial difficulties. A probabilistic tree modeling framework for estimating the overall traffic states is proposed in this paper. Firstly, we extract several typical traffic states covering the overall characteristics. Secondly, the state predicting algorithm based on dynamic Variable-order Markov Model and Genetic Algorithm is developed, where different date attributes were evaluated separately. Finally, the prediction model using probability tree with multiple prediction steps is verified. Experimental results using traffic speed data in Shanghai demonstrate the high accuracy and efficiency of the proposed method.