Research And Application Of Vehicle Trajectory And Checkpoint Traffic Flow Prediction Technology Based On Big Data Mining In Urban Road Network

In recent years,with the continuous improvement of sensor perception technology,Internet of Things communication technology and computer science,more and more technologies based on various types of big data are applied in the construction of smart transportation.The mining of historical traffic big data can be used to analyze future traffic trends and make corresponding countermeasures.For example,the prediction of vehicle trajectories can provide strong support for path planning and accurate recommendation;the prediction of short-term traffic flow plays an important role in Traffic command,congestion guidance,and police dispatch.However,for the increasingly perfect vehicle checkpoint system in urban road network,the traditional research methods based on GPS trajectory are no longer applicable.Therefore,this paper takes the passing data collected by the urban road network checkpoint system as the object,conducts a systematic prediction technology research,and develops the urban road network checkpoint prediction system based on the actual needs of enterprises,and verifies the effectiveness of the proposed method by using the real traffic data set.The main contents of this paper are as follows:In the first chapter,we expound the practical significance of traffic big data prediction,introduce the research background and technical status quo,and summarize the research status quo from three aspects:sequence and graph representation learning modeling,moving object trajectory prediction and short-term traffic flow prediction.Combined with the needs of the project and the shortcomings of the current research,this paper puts forward the research content and organizational structure of this paper.In the second chapter,based on representation learning,a method for calculating the similarity of traffic jams is proposed.In this method,firstly,we use word2vec to represent and learn the vehicle trajectory sequence flowing in the checkpoint,then we the road network checkpoint system as a checkpoint network diagram,and use the unsupervised representation learning method GraphSAGE to model its network connection topology.The combination of the two methods can obtain the low-dimensional dense vector representation of the checkpoint coordinates,and on this basis,we define the traffic checkpoint similarity,providing a theoretical basis for the follow-up study.Finally,the correctness and rationality of the representation learning model of the checkpoint proposed in this chapter are verified by the visualization of the T-SNE dimensionality reduction method and the example of the similarity degree of the traffic checkpoint.In the third chapter,a prediction model of vehicle checkpoint trajectory in urban road network is proposed.Firstly,the vehicle trajectory is mapped based on the checkpoint embedding vector obtained in Chapter 2,and the time sequence relationship in the vehicle trajectory sequence is modeled by long-term and short-term memory neural network,then the external factors that affect the vehicle trajectory selection are ed based on the traffic rules,which are embedded as auxiliary features to describe the trajectory,and both of them are input to the multi-layer perceptron classifier,and the output is not The prediction result and probability of the track position are obtained.Finally,the validity of the proposed model is verified by the real vehicle checkpoint trajectory data set collected in a city’s checkpoint system.In the fourth chapter,a short-term traffic flow prediction model for urban road network is proposed.In order to model the complex and dynamic spatiotemporal correlation between the traffic flow of the road network checkpoints,we design a deep neural network model based on spatiotemporal attention mechanism.In this model,a recurrent neural network unit based encoder-decoder architecture is adopted.Firstly,high similarity neighbor checkpoints are selected for the target checkpoint based on the similarity matrix of traffic checkpoint.A spatial attention mechanism is designed in the encoder to capture the impact of neighbor checkpoints traffic flow on the target checkpoint.In the decoder,a time attention mechanism is introduced to capture the temporal dependencies.Finally,the validity and interpretability of our method are verified by the real data set collected by a city’s checkpoint system.In the fifth chapter,based on the actual needs of enterprise projects and the three technical theories and methods proposed in this paper,we develop a system named Urban Road Network Checkpoints Prediction System,which provides the functions of checkpoints information query,traffic checkpoint similarity calculation,vehicle trajectory prediction and traffic flow prediction.Futhermore,we introduce the design ideas and technical routes of each functional module in detail.Finally,two prediction examples are used to verify the results availability and effectiveness of the system.The sixth chapter summarizes the research work and achievements of the paper,and looks forward to the future research work in the field of network checkpoint prediction.