The Prediction Method Empirical Research Of Urban Railway Transportation Flow

With the merits of high rate, large freight volume and comfortable, urban railway transit is eventually becoming the indispensable part of urban public transport system.In the field of urban railway transit passenger organization optimization, predicting short-term passenger volume has becoming the key technology. Combining the date of metro line one with railway characteristics, throughout this paper, statistical characteristics of railway passenger flow and the prediction models are researched,the main research content includes:(1) According to the length of prediction and accuracy requirements, passenger flow prediction is subdivided into three groups: the long-term prediction of traffic programming, the mid-term prediction for the traffic management and the short-term prediction for the traffic induction. The applicabilities and merits faults of macro prediction method and micro prediction model are analysed.(2) In the paper statistical features of urban rail transit passenger flow are analyzed. Including annual statistical characteristics, the main features are the cyclical nature, imbalance feature and highly nonlinear characteristic; Different to the normal passenger flow, the variation trend of holiday passenger flow is highly volatile. Based on the features of the passenger flow prediction models are divided into normal passenger flow and holiday passenger flow two types.(3) This paper established a prediction model based on Wavelet and Neural Network, this method is applied to the original passenger flow changes. This method firstly decomposes the original time series into different frequencies information by Wavelet, then Neural Network learn from the decomposed information and to forecast. Through the results of examples: in terms of iteration convergence accuracy the prediction model are improved greatly.(4) This article proposed the K-non-parametric regression method and it suit for to the change of the passenger flow during the holidays. Through learning the model principle, determine the algorithm step of model, and finally forecast the NationalDay of passenger and the Yuan Dan passenger flow. Results showed that the accuracy prediction of the holiday traffic is higher.