Research On The Short-term Traffic Flow Forecasting Method Of Based On Phase Space Reconstruction And SVR

Combining transport information with high-tech,such as cloud computing and the Internet of Things,has become one of the important measures for the development of smart cities.It plays a pivotal role in improving road capacity,solving urban road congestion,and reducing environmental pollution.Traffic flow forecasting is one of the key element of the information processing and intelligent control in ITS,and it is also one of the important application foundations for improving the level of traffic services and enhancing the user experience satisfaction.Because of the complexity and time-variation of traffic flow in urban roads,it is difficult to grasp the law of traffic flow changes and predict traffic flow conditions at any time,so it is of great significance for real-time prediction of short-term traffic flow.This paper takes the complex urban traffic network as the research background,applies the open microwave data provided by the OpenITS system,analyzes the characteristics of the collected traffic flow data,and obtains the periodic characteristics of the data presented by each detection point.Pearson's correlation coefficient was then used to measure the correlation between various microwave detection devices,and the correlation between the test points and the predicted points was verified based on the correlation results.Based on the study of phase space,this paper analyzes the necessity of traffic flow as chaotic time series.In the process of solving the reconstructed phase space,in order to adapt to the reconstruction conditions of the phase space,the conditions and criteria of the labeled sample data in k-nearest neighbors(KNN)are modified.Therefore,the basic parameters of the reconstruction phase space are obtained:delay time and embedding dimension.At the same time,this work belongs to the input module of short-term traffic flow prediction model based on phase space reconstruction and SVR,which provides the integration basis for the conversion of sample data to model input.In this paper,Support Vector Machine(SVM)is studied,and then Support Vector Regression(SVR),which is similar to SVM and applies to prediction problems,is applied.Due to the nonlinear characteristics of traffic flow data,Gaussian kernel functions are used to map the input data in high-dimensional space when the model is built.The process of constructing and solving this model is as follows.First,the phase space reconstruction theory is applied to integrate the sample data.Then,the reconstructed data is input into the SVR model to construct a short-term traffic flow prediction model based on phase space reconstruction and SVR.After that,train the model and use the more common ten-fold cross validation method to optimize the parameters.Finally,the model is tested by using the measured traffic flow data,and the traffic flow evaluation index is analyzed.At the same time,the model is compared with the neural network and the conventional SVR model.Through experimental verification,the model predicts better performance and can more effectively perform short-term traffic flow forecasting.