Research On Expressway Traffic Congestion Identification And Prediction Method

As one of the major transportation routes,the development and construction of expressway is inextricably linked to the growth of both the national economy and society.However,with the dramatic increase in the number of motor vehicles,the imbalance between supply and demand of highway traffic becomes more and more prominent.Traffic congestion identification and prediction is the core component of ITS,and the traditional traffic condition identifying and forecasting methods are not well adapted to the current complex and changeable road traffic system.Therefore,this paper conducts an in-depth study on the traffic congestion identification and prediction methods for highways,so as to effectively alleviate expressway congestion and enhance expressway operational efficacy.First,in order to select the indicators of the highway traffic state characterization parameters and to verify the reliability of the obtained traffic data,data research was conducted on the highway The traffic flow,average speed and traffic density were selected as congestion indicators and classified into four traffic states.Afterwards,the flow data was pre-processed and the dataset was constructed for the study of highway congestion identification and prediction.Then,an improved FCM highway traffic congestion identification algorithm was proposed to address the instability of the traditional FCM algorithm in randomly determining the initial clustering centers and the problem of ignoring the influence of the variability among the characteristic parameters congestion discriminative indexes on the calculation of the similarity measure among the samples.Firstly,the concept of feature weight vector and difference degree of samples was proposed,and the value of initial clustering center was improved by using weight difference degree fuzzy clustering algorithm,and then the improved algorithm of determining the weight of congestion discriminant parameter indexes in fuzzy clustering analysis from subjective and objective perspectives was proposed by using entropy value method and AHP combination algorithm.The improved FCM algorithm was compared and validated,and the results showed that the overall performance of the improved algorithm proposed in this paper was better,and the recognition accuracy of traffic state was above 0.95,which verified the feasibility of the proposed improved FCM highway expressway congestion recognition algorithm.Finally,to address the problem that most traditional highway congestion prediction methods constructed a single prediction model,and most of them were based on the temporal characteristics of traffic flow,lacking deep mining of the spatial characteristics of traffic flow,resulting in unsatisfactory prediction results,a highway traffic congestion prediction method based on SVR-LightGBM model was proposed.Firstly,based on the clustering results of the algorithm in the previous chapter,the specific value range of each discriminant index under different traffic states was obtained,and then the threshold value of congestion index under each state was calculated by quantifying the traffic states.Then the SVRLightGBM forecasting model was implemented to forecast the congestion index of the target road section by the upstream and downstream historical period traffic flow data.The model used MAE,RMSE and EC evaluation indexes to compare with the prediction results of other four combined and single models,and the results respectively reduced MAE by 42.53%,16.64%,44.52% and 40.68%,and RMSE by 37.10%,14.23%,37.96% and 24.09%.It indicated that the proposed SVR-LightGBM expressway traffic congestion prediction model proposed in this paper had high prediction accuracy,thus verifying the effectiveness and practicality of the method.