Research On Expressway Traffic State Prediction Based On Swarm Intelligence

In recent years,with the gradually improvement of highway network in China,the pressure of transportation in China has been regular relieved.However,traffic congestion is following,which cause the negative impact on development of China's social economy.Therefore,it is significanct to establish a intimately understanding of the current situation of China's highway and establish a traffic state prediction system,which is consistent with the characteristics of traffic flow data.According to the characteristics of traffic flow data,traffic state identification model is established and this paper consider direct and indirect perspectives to establish predict traffic state model.Firstly,traffic state identification and predict traffic state are taken as the research object in this thesis,the domestic and overseas research process are explained from the two aspects of traffic state identification and traffic flow prediction,deficiens of existing researches are summarized and the research contents and technical roadmap are proposed.Secondly,the traffic flow characteristics are analyzed with using the traffic flow data and the time correlation analysis and spatial correlation analysis are completed respectively simultaneously.Combined with the analysis results,the traffic flow anomaly data identification method is proposed.The abnormal data is divided into random data to be repaired and continuous data to be repaired.The data repair methods of average value filling and historical data weighting value filling are designed respectively to provide data basis for the next traffic state identification and prediction.Thirdly,the domestic and overseas classification of traffic state are analyzed,and the traffic state identification indicators are determined including the volume,speed and time occupancy.This paper proposes a traffic state recognition algorithm based on swarm intelligence optimization.The results show that compared with the traditional fuzzy c-means clustering algorithm,the algorithm proposed in this paper can achieve faster convergence and have a productive recognition effect.Finally,with introducing the deep learning theory,the direct and indirect traffic state prediction models based on GRU network are designed,and analyzes the performance of the two algorithms combined with the traffic flow data of Beijing Hong Kong Macao expressway.The results show that prediction accuracy of the two methods all better than 88%.Moreover,the direct prediction algorithm shows better prediction effect,and the prediction accuracy is 2.1% higher than the indirect prediction algorithm.