The Research Of Short-Term Traffic Flow Prediction Based On The Combination Method Of Time Series Decomposition

Research on the theory and technology of short-term traffic flow prediction and its practical application are of great significance to alleviate the increasingly serious traffic problems,boost the efficiency of urban traffic system,improve the urban intelligent transportation system(ITS),the development of virtual urban transportation system.This paper studies the problem of short-term traffic flow prediction from the aspects of theoretical technology,prediction model,experimental test and practical application.Firstly,this paper explores the traffic data collection,traffic flow prediction principle,missing data completion,abnormal data processing technologies.Especially for the abnormal varieties in the traffic flow time series,an abnormal data correction technology based on normal distribution is proposed and refined.These studies provides technical foreshadowing for the prediction model research.Secondly,the mathematical principles of traffic flow prediction models commonly used in existing studies are explored.By refining the periodic-trend decomposition(PTD)and singular spectrum analysis(SSA),the traffic flow time series is decomposed into three components:trend,period and residual,and the potential change characteristics of traffic flow are analyzed.On this basis,a traffic flow prediction model based on combined time series decomposition is proposed to improve the prediction accuracy and the stability of multi-step prediction.Based on the traffic flow data of interstate highway in Minnesota,in U.S.,the paper discusses and analyzes the prediction results of different detector stations,the stability of multi-step prediction and the running time of the model,and then summarizes the advantages and disadvantages of each test model.Finally,based on the simulation platform of virtual urban traffic system,the application program of traffic flow prediction executable program is designed and developed.Based on the traffic data from vehicle detection station of Chongqing,the case analysis of Chongqing road network traffic flow prediction is completed,and the theory of traffic flow prediction technology is put into practice.