Research On EEMD-based Multi-component Combined Short-term Traffic Flow Forecasting Method

With the economic development and social progress,the number of vehicles in China has increased exponentially,and the problem of traffic congestion has become more and more serious,which has caused unnecessary losses to our country economic development.In order to alleviate traffic congestion,intelligent transportation systems are more and more widely used in urban construction,and short-term traffic flow prediction is an important part of intelligent transportation systems,so its development has important significance for improving the efficiency of the transportation system.The real-time and accuracy of short-term traffic flow prediction determine the control accuracy of traffic flow and the effect of guiding traffic flow,and its research has become one of the important contents of research in the field of traffic.After nearly ten years of development,short-term traffic flow forecasting has formed a variety of forecasting methods,some of which are very complete and are actually used in actual traffic systems.Because traffic flow is easily affected by weather,historical traffic flow,and human factors,traffic flow has the characteristics of randomness and nonlinearity.Using a single prediction method to predict traffic flow cannot meet the needs of modern intelligent transportation systems.According to the advantages and disadvantages of existing traffic flow prediction methods,this thesis proposes a multi-component combination model based on ensemble empirical mode decomposition(EEMD)to predict short-term traffic flow.First,the EEMD decomposition method is used to decompose the original traffic flow time series,according to the waveform characteristics of the decomposed component sequence,it is divided into high frequency components,intermediate frequency components,and low frequency components.Then use wavelet neural network model,least squares support vector machine model,Autoregressive Integrated Moving Average model to predict high frequency components,intermediate frequency components and low frequency components,the genetic particle swarm optimization algorithm is used to select the parameters of the wavelet neural network model,and the particle swarm optimization algorithm is used to select the least squares support vector machine model parameters,the purpose is to improve the prediction accuracy of the two models.Finally,the least squares support vector machine model is used to superimpose the predicted value of each component to obtain the final predicted value of traffic flow.Collect traffic flow data of New York Huguenot Ave and New York Stafford Huguenot Ave as data samples,establishing the above prediction model.Input the data samples collected in the two places to the established model for traffic flow prediction.The prediction results are compared with a single model(least squares support vector machine,Autoregressive Integrated Moving Average,wavelet neural network)using two evaluation methods,root mean square error and average percentage error,the comparison shows that the prediction model proposed in this thesis has higher prediction accuracy than other single models.