Traffic Flow Prediction And Congestion State Assessment Based On Temporal And Spatial Characteristics

With the rapid development of Internet of things,cloud computing,big data and artificial intelligence,the application of deep learning is becoming more and more popular.In this paper,a traffic state prediction model based on deep learning is established,which takes the bus as the floating car,the GPS data on the bus as the research object,and the speed as the main characteristic parameter.In recent years,graph neural network(GNN)is emerging,which can provide new ideas for the construction of road network spatial dependence model.In this context,this paper combines the spatial and temporal dependence modeling,and establishes the traffic flow prediction model based on the spatial-temporal characteristics.The specific research contents are as follows: firstly,this paper introduces the bus GPS data and weather data as well as the acquisition methods,illustrates the feasibility and advantages of using bus GPS data for prediction,and the influence of weather factors on traffic flow.Then the development of deep learning and the neural network algorithm used in this paper are described in detail.Next,the spatial parameters of road network traffic flow extracted based on graph attention neural network are taken as the characteristics,and the traffic flow prediction model is constructed by using the time dependence of the cyclic neural network which increases attention mechanism on road network traffic flow.Finally,the sequence to sequence model is used to predict the traffic flow,and the automatic coding mechanism is introduced to reduce the complexity of the model in the prediction,and the planned sampling mechanism is introduced to improve the prediction accuracy in the training.Finally,an optimized traffic flow prediction model is proposed.Most of the researches on traffic flow prediction are based on flow parameters.The reason why this paper chooses speed parameters for prediction is that speed parameters have the characteristics of large mutation and large fluctuation in a short time relative to flow parameters,which can better reflect the short-term traffic state.Therefore,this paper explores the method based on speed parameters,which provides a new idea for traffic state recognition based on public transit GPS data.At the same time,in order to reflect the reality more objectively,this paper also considers the weather information data corresponding to GPS data in the data set.The experimental results show that the model has a good prediction effect in different degrees on the three indicators,and has a certain improvement in the accuracy and structure simplicity.