Research On Traffic State Identification And Prediction Based On Multi-Feature Fusion

With the rapid development of the national economy,and the number of private cars in my country has increased year by year,and the resulting traffic congestion has become more and more serious.It is necessary to establish a sound traffic management policy,and therefore,an intelligent transportation system is required to rationally allocate traffic resources.By predicting the traffic operation status,it can not only provide information for the traffic management department,but also provide great convenience for the public to travel.The thesis mainly focuses on the identification and prediction of traffic operation status,explores the factors that affect the traffic operation status,and designs a multi-feature fusion traffic prediction model based on the influencing factors.First of all,this article takes the judgment of traffic operation status and the prediction of traffic operation status as the research object,analyzes the current research situation at home and abroad,summarizes the existing achievements and shortcomings,and proposes the research content and technical route of this article on this basis.Secondly,explore the influencing factors of the road operation status,including the impact of weather,major events,and the temporal and spatial characteristics of the traffic flow on the road operation status.The changes in the traffic operation index and the average speed reflect the impact of different rainfall,snowfall and haze weather conditions on road traffic,and the impact on road traffic operation is analyzed from the changes in the traffic congestion index when a major event occurs.Analyze the time correlation and spatial correlation that affect traffic flow,analyze the daily and weekly traffic flow changes,and use the Pearson correlation coefficient to verify its correlation,proving the similarity and cycle in the time dimension The similarity analysis of the upstream and downstream traffic indexes of four adjacent road sections shows that the traffic flow has a certain correlation in the spatial dimension.In view of the existing traffic prediction models that rarely consider multi-feature fusion prediction methods,this paper proposes a traffic prediction model based on multi-feature fusion.First,the attention mechanism and GRU are used to extract the time characteristics of the road,and GCN is used to extract the spatial features.Integrating the temporal and spatial features and adding periodic factors and weather factors to further optimize the model,finally get the MFSTGCN traffic prediction model proposed in this paper.The traffic flow and speed are predicted on the real traffic data sets Pe MSD4 and Pe MSD8 and compared with other models.In contrast,using MAE and RMSE as evaluation indicators,the results show that the model proposed in this paper has the best effect on each indicator.Whether it is short-term or longterm prediction,the stability and error of the model are better than other algorithms.Finally,based on the above-mentioned research,this article discriminates and predicts the state of traffic operation.First,the K-means clustering algorithm is used to cluster the traffic flow and speed,and it is judged as unblocked,basically unblocked,lightly congested,and severely congested.Then use the multi-feature fusion traffic prediction model to predict the traffic volume and speed of a week,and then perform cluster analysis on the predicted results again to determine the traffic operation status,compare with the previous discrimination results,and use the confusion matrix to determine the accuracy of the discrimination.After evaluation,the overall accuracy rate is 91.7%,which proves that the method adopted in this paper can make accurate judgment and prediction of traffic operation status.