Early Warning Analysis Of Expressway Traffic Flow Based On Graph Neural Network

Expressway has been essential in our daily life,but with the development of society,traffic jams are often occured on the expressway.Early warning of traffic flow can effectively help the expressway operator to make relevant treatment in time and provide sufficient emergency time for relevant departments.Due to the complex expressway networks and obvious periodicity of traffic flow,the reasonable application of Graph Neural Network can effectively improve the accuracy of early warning.Based on Graph Convolution Neural Network and Long Short Term Memory Neural Network,Expection-Maximization Algorithm and Gaussian Mixture Model,this paper constructs an expressway traffic flow early warning model and evaluates and tests the model with real data.Firstly,according to data which record vehicle passing on the expressway,the characteristics of traffic flow are analyzed.The data shows that the traffic flow of expressway has a strong periodicity,and there is a strong correlation between different locations.Therefore,this paper proposes a Graph Convolution-Long Short Term Memory Neural Network to predict the traffic flow,and use the Adaptive Moment Estimation Algorithm to optimize the parameters.The model can not only make full use of the periodicity of the data,but also effectively consider the influence relationship between the traffic flow at different locations.After obtaining the prediction results of traffic flow,the Gaussian Mixture Model is used to analyze the maximum threshold of normal traffic flow at each node at each time,and the Expectation-Maximization Algorithm is used to solve the parameters.Finally,the prediction result is brought in to judge whether it is an abnormal situation and give early warning information.The model is tested by using two groups of traffic volume data of an expressway under different conditions,and compared with other common time series models.According to the evaluation indexes goodness of fit,Root Mean Square Error and Mean Absolute Error calculated by different models,the Graph Convolution Long Short Term Memory Neural Network can shows better prediction results than other models.On the other hand,the introduction of spatial information can effectively improve the accuracy of traffic flow prediction model.Then input the prediction results into the reasonable threshold of traffic flow determined by Gaussian Mixture Model for judgment.After comparing with the results of real value judgment,it can be found that the accuracy of judgment is very high,that is,it ca n provide reliable early warning information.