Research On Urban Expressway Traffic State Recognition And Prediction Based On Multi-source Data

Urban expressways are an important part of urban transportation and affect the operational efficiency of urban transportation systems.With the improvement of multi-source traffic data collection technology,the identification and prediction of urban expressway traffic status fusion and multi-source data is the basis for traffic management control,which not only helps the traffic management department to accurately grasp the urban traffic operation situation,but also helps For travelers to choose a reasonable travel route,improve the efficiency of urban traffic operation.However,domestic and foreign scholars still have deficiencies in their research,mainly as follows: First,fuzzy mean clustering algorithm is usually used to divide the traffic state,and the effect of fuzzy mean clustering is greatly affected by the selection of initial values;second,many The research of source traffic data fusion is mainly in the data layer and decision layer,and there is little research on the fusion of the feature layer;the third is the use of deep learning algorithms for traffic state recognition and prediction,because there are many algorithm parameters,it is difficult to find the optimal Parameter combination.Based on this,this paper extracts the characteristics of multi-source traffic data to obtain traffic flow and travel speed,introduces simulated annealing particle swarm optimization to optimize the initial selection of fuzzy mean clustering,and uses Bayesian algorithm to optimize the hyperparameters of Xgboost traffic recognition algorithm The genetic algorithm is used to optimize the hyperparameters of the Bi LSTM traffic data prediction algorithm,and the predicted traffic flow and travel speed are input into the traffic state recognition model to obtain the predicted traffic state.The main research contents of this article are as follows:(1)Extract traffic features from traffic bayonet data to obtain traffic flow,and extract traffic features from floating car trajectory data to obtain travel speed.Recognize and repair abnormal data,and use the spatial-temporal correlation theory to select relevant road sections as the basis of data fusion research.(2)Using simulated annealing particle swarm optimization fuzzy mean clustering(SAPSO-FCM)initial clustering center to establish a traffic state division model to achieve a scientific division of traffic state.(3)Using Bayesian optimization of the hyperparameters of the Xgboost algorithm to form a BO-Xgboost traffic state recognition model,to achieve Xgboost hyperparameter rapid optimization,and to establish a traffic state recognition model.(4)The genetic algorithm is used to optimize the Bi LSTM hyperparameter combination,which is applied to the prediction of traffic flow and travel time to achieve highprecision prediction of traffic data,and the predicted value is input into the BO-Xgboost model to obtain the predicted traffic state.The example shows that the final value of the objective function of the SAPSO-FCM algorithm is always the same,and is lower than the final value of the objective function of the FCM algorithm,which has better convergence and stability.Compared with the control model,the average accuracy of the BO-Xgboost model is 99.65%,with the highest average accuracy.GA-Bi LSTM model shows higher prediction accuracy in traffic flow and travel time prediction,and the accuracy of traffic state prediction reaches 90.97%.In this paper,research is conducted on the division,recognition and prediction of urban expressway traffic status under multisource data to form a set of technical solutions that integrate traffic flow and travel speed,which enriches the research results in the field of traffic status.