Urban Railway Traffic Passenger Flow Short-term Forecasting Method And Empirical Research

With the constant expansion of the city, the continuous growth of the population and the popularization of private car, traffic congestion、environment pollution、energy consumption and other problems are getting more and more serious, which run the opposite of the wishes that people are free to travel with, the city facing huge traffic pressure. Timely、environmental and comfortable urban railway traffic with big freight volume and good benefit has become the main developing direction to solve these traffic problems. Passenger flow volume is the base of urban railway traffic network planning and designing, while the forecast of passenger flow is the key technology of urban railway traffic project construction.Around the problems of railway traffic passenger flow short-term forecasting, the research work of the paper is as follows:Firstly, temporal characteristics and periodicl laws are obtained by statistics and analysis of passenger flow volume during tomb-sweeping day in Beijing city, and the mainly influential factors are obtained as well. In addition, data sample types are effectively classified through clustering analysis of line passenger flow and arriving passenger flow, which provides reliable data support for the improved model and empirical research.Secondly, according to the limitation structure status of RBF neural network model, a modified weighted neural network forecasting model with a plurality of modules based on temporal characteristics is established. And the forecasting example of railway traffic passenger flow proves that the improved forecasting model has higher prediction accuracy.Thirdly, according to the limitation of single kernel function, an improved support vector machine forecasting model based on mixed kernel function is established. And the forecasting example of arriving passenger flow proves that the improved forecasting model has better fitting effect, which shows feasibility in the field of short-term passenger flow forecasting.Finally, considering the advantages of RBF neural network and support vector machine in learning methods、modeling methods and structural characteristics, the two prediction methods will be combined together. Then a combination forecasting model based on gray relational degree maximum is established. By the forecasting example of line passenger flow during tomb-sweeping day, the combination forecasting method obtains greater relevance than single forecasting methods, proving that the combined forecasting model is superior to single forecasting methods.