Hybrid Process Neural Network based on Spatio-Temporal Similarities for Short-Term Traffic Flow Prediction

Spatio-temporal similarities, one of the characteristics to describe the relativity of traffic phenomenon, can be utilized to predict short-term traffic flow. These similarities not always appear at spatial adjacent road links because of complexity of road network. In this paper, we adopt Cross-Correlation Function to depict similarities between different traffic flow series according to the observed flow data. The process characteristic generalizes the evolvement rules of traffic flow which are essentials need to be tackled by a prediction model. After choosing the most correlative road links and their time delay instead of the upstream or downstream ones, a Hybrid Process Neural Network is constructed to predict shortterm traffic flow, which uses various scales to catch traffic features such as daily-periodicity, weekly-periodicity and spatiotemporal process, since a simple model is not good enough to depict all these rules. Application of the proposed method is demonstrated, and the experimental results show that our method outperforms other compared methods.