Traffic Flow Prediction And Traffic State Research Considering Temporal And Spatial Characteristics

With the continuous development of urbanization,the number of people who choose private transportation to travel is increasing.With limited road resources,the increasing number of vehicles will inevitably cause a series of traffic problems,among which traffic congestion is a problem faced by many countries.Traffic flow prediction can provide effective data support to alleviate traffic congestion,and more accurate and intuitive traffic flow information can enable traffic participants to make more suitable decisions.In order to effectively alleviate the problems caused by traffic congestion,this paper constructs a traffic flow prediction model based on deep learning from the perspective of traffic flow prediction,with the aim of improving the prediction accuracy of traffic flow.In order to more intuitively indicate the traffic status on the road section,a traffic status prediction model based on deep learning clustering algorithm is further proposed on the basis of traffic flow prediction.The details of the work are as follows:(1)The influencing factors of traffic flow are analyzed,and traffic flow,average speed and time occupancy are selected as the characteristic parameters to measure the traffic flow changes.The traffic data are visualized in the time dimension and spatial dimension,respectively,to analyze the relevant characteristics of traffic flow,and introduce the relevant deep learning theoretical algorithm models to provide the theoretical basis for establishing traffic flow prediction models.(2)Construction of a spatio-temporal attention traffic flow prediction model based on collaborative approach.A graphical convolutional network and a temporal convolutional network are introduced to extract the variation characteristics of traffic flow in spatial and temporal dimensions.A spatio-temporal attention mechanism based on a collaborative approach is proposed to consider the potential correlation between traffic flow in temporal and spatial dimensions,which is used to dynamically capture the long-term dependence between spatio-temporal information of traffic flow at the same time,and to clarify the influence of different feature information in different spatio-temporal dimensions.A new activation function is introduced to improve the prediction performance of the traffic flow prediction model.(3)Application of a traffic flow prediction model with collaborative spatio-temporal attention.The proposed traffic flow prediction model is applied to two real datasets Pe MSD4 and Pe MSD8 to validate and analyze the prediction performance of the model.First,the data are pre-processed,optimized and segmented.Secondly,the model evaluation indexes and hyperparameter settings are determined.Finally,the accuracy and timeliness of the model are demonstrated by comparing and analyzing with existing prediction models in terms of both accuracy and timeliness of traffic flow prediction.(4)Traffic state prediction model construction and validation of spatio-temporal attention clustering algorithm.Traffic flow,average speed and time occupancy are used as the basis for measuring traffic state classification.According to the operational characteristics of highway traffic flow,the fuzzy c-mean algorithm and Gaussian mixture function in the soft clustering method are used to classify the sample data into three categories: smooth,stable and congested,and the indirect prediction of traffic state is realized by combining the spatio-temporal attention traffic flow prediction model based on the collaborative approach proposed in this paper.The prediction results of the model are analyzed and verified by the Pe MSD8 dataset,and the statistical accuracy can reach 91.4%.