Short-Term Passenger Flow Prediction of a Passageway in a Subway Station Using Time Space Correlations Between Multi Sites

Short-term passenger flow prediction of urban rail transit is very important to traffic management. The prediction result is a key input for the crowd early warning system. However, a reliable approach of short-term passenger flow prediction is still greatly desired. At present, most researches focus on single-site prediction (SSP), and its technique cannot be improved due to its maturity, and any further breakthrough is difficult to make. In this paper, a multi-sites prediction method (MSP) of passenger flow in subway station is proposed. Real time passenger flow data collected from multi-sites in a subway station is used as inputs, and delay parameter is introduced to identify the correlation between measurements at multiple sites in this paper. In order to achieve a stable predictive effect, wavelet decomposition and reconstruction are used to process the data. Dynamic weights combination of Support Vector Machines (SVM) and Radial Basis Function Neural Network (RBF) are applied to obtain the prediction of each frequency component. A case study of a station in Beijing subway shows that the proposed MSP method reduces the average absolute prediction error by 15.8% compared with the SSP. It is concluded that the proposed method has good performance in the case.