Short-term Traffic Flow Prediction Based on Multi-Auxiliary Information

Aiming at the problem of short-term traffic flow prediction, on the basis of considering the dynamic change of traffic flow diffusion, this paper proposes a spatio-temporal prediction model that integrates multiple-auxiliary information, and carries out spatio-temporal modeling for the traffic flow state of a single detector, and then realizes the forecast goal of traffic flow. In this paper, firstly, based on the PageRank algorithm, the RoadRank algorithm is proposed to obtain the real-time traffic flow diffusion characteristics of the road network. Secondly, feature construction of multiple auxiliary information such as time, period, weather, etc., and on the basis of the above multiple auxiliary features, combined with the Encoder-Decoder framework that considers the attention mechanism, a spatio-temporal prediction model framework for multi-auxiliary information fusion is given. Based on METR-LA traffic dataset and NOAA weather dataset, a case study is carried out to verify effectiveness of multi-auxiliary features and superiority of the hybrid model.