Research On Short-term Forecast Of Railway Passenger Flow Based On The Passenger Flow Excited Levels Model

Railway passengers flow forecasting is an important basis of passenger traffic organizing and an effective way for the passenger traffic enterprise to grasp the trends and demand of transportation market development.This paper focuses on short-term passenger volume forecasting problem between two cities. The objective is to investigate appropriate forecasting methods in order to guarantee the precision.This paper begins the research work from observing the fluctuation characteristics of short-term passenger railway volume between origination-city and destination-city, and finds some special properties of the volume change. On this basis, the paper gives a conclusion that methods based on combining time series analyzing and qualitative approach should be used in short-term passenger volume forecasting.Short-term Railway Passenger Flow Forecasting method Based on Passenger Flow Excited Levels Model is enlightened by Boer's atomic-model in physics. The model include two key parts, the one is evaluation levels and the increment in excited states, the other one is time series forecasting in ground states.In the study of evaluation levels and increment in excited states, the factors of passenger flow transition is analyzed at first, then Fuzzy comprehensive evaluation is applied to evaluate these factors. To overcome the shortage of historical data ,the increment of learning samples are got by clustering analysis the time series data from Ticket sale record.In the study of time series forecasting in ground states, the method for recognition and processing singular values is proposed, then LS-SVM is applied to forecast. In order to improve efficiency and generalization ability of forecasting, genetic algorithm is applied to select parameters of LS-SVM.Finally, the method is applied to the actual passenger flow volume daily data of Guangzhou to Wuhan. The model and methods are effective by the test of example.