Research On Average Speed Prediction And Congestion Identification Of Bus Lanes On Urban Trunk Road

Due to the rapid growth of residents’ travel demand,traffic congestion on urban trunk roads is becoming more and more serious,which has become a major obstacle in the development of today’s cities.The development of urban public transport provides a new way for relieving traffic congestion.Under the background of the implementation of the public transport priority strategy,using the GPS data of buses to predict the average speed of buses on the bus lanes,and then judging the running state of buses,can be an important means to improve the efficiency of urban public transport network.Based on this,the research content of this paper is as follows:Firstly,data preprocessing and feature extraction based on the GPS data of bus.After preliminary processing of the original bus GPS data set,the target trunk road is divided according to the method of isometric division and node division,and then the characterization parameters of traffic state are extracted.At the same time,considering that the influence of road environment and other factors on traffic state parameters is rarely considered in current studies,the influencing factors of traffic state parameters are extracted from four aspects: weather,time,road conditions and bus operating environment.And spearman correlation analysis is used to eliminate the influencing factors with collinearity.Secondly,the prediction model of road section average speed considering temporal and spatial characteristics is constructed.By combining he advantage of Convolutional Neural Network(CNN)in extracting spatial information and the ability of Bidirectional Long-term Memory Network(BiLSTM)in predicting temporal data,the Sparrow Search Algorithm(SSA)is used to optimize the parameters of the neural network,and finally the average speed prediction model of SSA-CNn-Bilstm-Attention section is constructed.Then,through the analysis of the importance of the influencing factors,the factors that have an important impact on the prediction of the average speed of a section,such as weekend,time period,pedestrian three-dimensional street crossing facilities,the number of bus stops,the number of lanes,commercial areas and schools,are input into the model for prediction.Through the evaluation and comparison of the prediction effect of the model,the validity of the prediction method is verified.Thirdly,the classification threshold of traffic congestion is determined based on clustering algorithm.For the defect of Fuzzy C-means(FCM)clustering algorithms,which tend to fall into local optimality,Genetic algorithms(GA)and Simulated Annealing(SA)algorithms have been combined to optimize the FCM clustering algorithms.Then the optimized FCM is used to cluster the average speed of road sections,and the traffic congestion is divided into three levels:smooth,normal and congestion.And the clustering results show that compared with the traditional FCM algorithm,the optimized FCM algorithm has better clustering effect,which has the advantages of better convergence speed and better stability.The final results show that the traffic congestion identification accuracy of equidistant sections is higher,reaching 89.4%.