Study On Short-term Traffic Flow Forecasting Based On Multiple Phase Space And Road Similarity

Traffic congestion has seriously affected people's travel and reduced the service capacity of roads.The best route for people to travel can be provided by short-term traffic flow forecasting,so as to relieve traffic congestion.Therefore,it is of great research significance and practical value to improve the prediction accuracy and prediction speed of short-term traffic flow.It is found that the problem of low prediction accuracy is existed in the short-term traffic flow forecasting model,and the prediction model is forecasted in the singlemachine environment,which leads to the slowdown of the forecasting model.Aiming at the above problems,a short-term traffic flow forecasting model is proposed in this paper based on multi-phase space and similarity degree of road,and the validity of the proposed model is validated in the end.The main research work and conclusions of this paper are listed as follows:Firstly,Aiming to the problem that the prediction accuracy is not high by using single phase space in chaos theory,a prediction algorithm based on multiple phase space is proposed.The average mutual information method is used to solve the delay time set and obtains the corresponding weight set in the algorithm.At the same time,the corresponding set of embedding dimension is computed to reconstruct the multiphase space,and the short-term traffic flow forecasting model is established with the KNN algorithm,which can well improve forecast accuracy.Secondly,a prediction algorithm based on multiple phase space and similarity of link is proposed to solve the problem that the prediction precision is not high for sparsely connected data.The algorithm takes the effect of road segment similarity into account on forecasting traffic flow.The positive and negative sets are calculated by measuring the positive similarity and negative similarity of the two sections,and the positive and negative correlation sections are combined.Spatial traffic flow forecasting model is established to further improve the prediction accuracy.Thirdly,the experiment is carried out by using multiple sets of GPS data of a section of road network.The experimental results show that the average percentage of relative error predicted by the short-term traffic flow forecasting model based on multiphase space and link similarity is reduced by 3.94%.The average absolute error and root mean square error of the two evaluation criteria are reduced ranging from 4.1 to 4.3,the prediction precision is improved obviously.When the data set reaches 80,000,the prediction model proposed in this paper is 15.74 times faster in the Spark environment than in the single machine environment,and the acceleration will increase with the increasing of the data volume,which has greatly shortened the model's forecasting time,and improved the prediction speed of the model.