Research On Prediction Of Freeway Traffic Flow In Holidays Based On EMD And GS-SVM

As a modern transportation infrastructure, highway is characterized by high speed. However, free of charge policy during holidays has caused traffic congestion that not only seriously affects the convenience and comfort for public travel, but also brings severe challenges for national traffic safety. Therefore, a scientific and effective traffic flow forecasting method and a simulation analysis of running state during holidays can effectively alleviate traffic congestion on highways during holidays in China, ensure safe and comfortable travel for the public during holidays and bring more social benefits, so as to promote the sustainable harmonious and sound development of highway traffic system of China.At first, this paper summarizes characteristics of highway traffic flow and the purpose of the public’s trip during holidays, analyzes factors influencing holiday trip and distribution characteristics of holiday traffic flow in terms of time, space and trip mode.While analyzing advantages and disadvantages of existing ways to forecast highway traffic flow during holidays, the paper pre-processes the holiday traffic flow data of a highway in Henan province from 2012 to 2015, uses empirical mode decomposition method to isolate noise data and make smoothing filtering, uses grid optimization to optimize regression parameters of support vector machine, predicts holiday traffic flow, and corrects weather influence factors according to the results of prediction. Comparative analysis is made between the results and that of common traffic forecast model to verify the accuracy and adaptability of the fusion algorithm.Finally, a holiday forecast model is built by TransModeler simulation software and the holiday highway running state is analyzed. Highway service level at different threshold values and the corresponding traffic flow are given, providing reference for highway operation and management.Research results show that in traffic flow forecast, fusion algorithm of empirical mode decomposition, smoothing filtering and grid optimization of support vector machine proposed in the paper is more accurate than non-decomposition method and the forecast results are more accurate than common prediction models. Simulation results of holiday model based on the TransModeler simulation software is more accurate, providing reference for the highway operation and management.