Research On Traffic State Analysis And Application For Urban Road Network Based On Data-Driven

Since China’s economy has entered the fast lane of development,the scale of cities has gradually expanded,the living standard of residents has significantly improved,the total number of motor vehicles has increased day by day,and the traffic congestion problem has become increasingly serious.Integrating the basic theoretical knowledge of traffic,mathematics,computer and other disciplines to study the traffic state characteristics of road networks and their evolution trends,and to explore the potential relevance in traffic networks,not only helps to improve the traffic service level of road networks,but also can provide theoretical support and new solution ideas for traffic congestion management.This thesis focuses on the research on the study of the traffic state in road network,which mainly includes road network traffic state classification,correlation analysis and traffic situation change analysis.First the traffic flow data of Beijing’s road network is introduced,and raw data noise is reduced by data cleaning and other techniques to ensure the reliability of the data.After data preprocessing,we conduct preliminary exploratory analysis to lay the foundation for our research.Then,The data of all road sections at the same time are used as a representation vector of the road network state,the best number of classifications is determined by multiple-indicator evaluation.K-means algorithm is used to classify the traffic state of Beijing’s road network.The classified traffic states are analyzed in daily changes on working and non-working days,the results show that the classification results can effectively represent the road network traffic state at different times.The road network traffic state is comprehensively analyzed from two aspects:traffic relevance and traffic situation changes.The correlation analysis is based on the theory of spatial auto-correlation.Firstly,the adjacency matrix is generated based on the adjacency relationship with road segments as nodes.To characterize the spatial relationship of road segments in the road network,then the reachability is introduced to construct a spatial weight matrix based on the shortest path distance.Finally,the temporal and spatial correlation characteristics of the road network is explored by spatial auto-correlation theories.In order to analyze the evolution of the traffic situation,the graph clustering method is used to realize the multi-level division of the traffic road network.Compared with the traditional similarity matrix construction method,the proposed method of constructing the similarity matrix based on distance-speed can aggregate nodes similar in traffic state and ensure spatial compactness.Based on the result of the preliminary division,spatial constraints are introduced as the termination condition to divide the road network again.The experimental results show that the results of multi-level division based on different traffic states can well reveal the changing rules and internal characteristics of the road network,which have guiding significance for the coordinated control of road network traffic and the identification of key road sections.There are 29 pictures15 tables,and 51 references in the body.