Short-term Traffic State Analysis And Prediction Considering Spatial-Temporal Information

Automobile society inevitably leads to environmental pollution,waste of resources and traffic congestion problems.How to effectively solve traffic congestion bottleneck has become a major issue faced by most of big cities.Accurate and real-time traffic state information can provide guiding advices for citizens' travel and traffic management.To analyze and forecast the short-term traffic state is a powerful measure to relieve urban traffic congestion,which has great research significance and practical economic benefits.Aiming at how to effectively solve the urban traffic congestion situation,as well as providing short-term traffic travel information for citizens in time,we propose a novel short-term traffic state analysis and prediction approach considering spatial-temporal information in this paper.Firstly,we propose the short-term traffic flow prediction algorithm based on optimized Adaptive Multi-kernel Support Vector Machine(AMSVM).Secondly,we consider the spatial-temporal correlation of traffic flow,and fuse the AMSVM's predicted value and spatial-temporal correlation predicted value with different weights.Besides,we utilize the Fuzzy C-means clustering(FCM)and Random Forest methods to analyze the feature parameters of traffic flow,and then obtain the predicted results of short-term traffic state.We carry out three main research works in this paper.1.Short-term traffic flow prediction based on optimized AMSVM.Firstly,we explore both the nonlinearity and randomness characteristic of traffic flow,and propose the AMSVM which hybridizes the Gaussian kernel function and polynomial kernel function.Secondly,we propose the Adaptive Particle Swarm Optimization algorithm(APSO)to optimize the parameters of AMSVM.Besides,we describe the detail processes about how to forecast short-term traffic flow according to the historical and real-time data.The proposed APSO-AMSVM prediction algorithm improves the prediction accuracy and efficiency at the same time,which can better adapt to the complex and dynamic traffic conditions,and realize the timely and high efficient prediction of short-term traffic flow.2.Short-term traffic flow prediction with spatial-temporal correlation.Firstly,we explore the temporal correlation of the distant historical traffic data,and obtain the historical temporal correlation predicted value.Secondly,we explore the spatial correlation of the spatial correlative points'traffic data,and obtain the spatial correlation predicted value.Besides,we fuse the AMSVM's predicted value,historical temporal correlation predicted value and spatial correlation predicted value with different weights,and then obtain the final predicted short-term traffic flow of the point of interest.The proposed short-term traffic flow prediction algorithm with spatial-temporal correlation(AMSVM-STC)further enhances the reliability and accuracy of the prediction results of AMSVM.It can fully mine the traffic flow information from several different dimensions,and can adapt to the extremely complicated urban traffic conditions and achieve a considerable predicted result.3.Short-term traffic state analysis and prediction.Firstly,we utilize the AMSVM-STC algorithm to predict the short-term traffic volume,speed and occupancy of different scenarios,respectively,and obtain the complete predicted values of traffic flow parameters.Secondly,we employ the FCM algorithm to obtain the historical traffic state information.Finally,we employ the Random Forest algorithm to analyze the predicted short-term traffic flow parameters,and then predict the short-term traffic state of the point of interest.The experimental results demonstrate that the evaluation approach of historical traffic state based on FCM is suitable for both freeway and urban road scenario,the Random Forest has higher prediction accuracy and generalization performance than other common machine learning methods.The experimental results accord with the general regularity of actual traffic conditions,so that it can provide the short-term traffic travel information for users timely and effectively.