Analysis And Prediction Of OD-Betweenness Centrality In Road Networks

As cities continue to grow and expand,urban road networks become increasingly complex,leading to worsening traffic congestion and necessitating effective traffic management and planning strategies.Traditional traffic forecasting management focuses on the traffic conditions of individual urban road sections,making it difficult to manage and predict from a more global perspective.Utilizing network centrality for global management is one approach,but the betweenness centrality(BC)based on the shortest path throughput,which has been applied to road network analysis,has certain limitations.For example,it assumes that the weights of all node pairs are the same and does not incorporate actual origin-destination(OD)information,making it challenging to adequately represent the complexity of urban traffic flow.To address these shortcomings,this paper introduces the origin-destination betweenness centrality(ODBC),a new centrality measure that combines actual OD flow information and modifies the betweenness centrality.Both theoretical derivations and experimental analyses show that it is positively correlated with actual traffic flow.This paper also proposes a feasible calculation method for ODBC,visualizes the results,and confirms its correlation with traffic flow.Furthermore,to analyze and manage the global road network more comprehensively,this paper plans to extract more information from historical ODBC data.Areas with higher ODBC values have higher traffic flow and higher social value in traffic guidance,navigation planning,and commercial billboard layout.These areas also have relatively high rankings in the road network,as we focus on identifying areas with high ODBC rankings in the next moment,transforming this issue into a ranking prediction problem.To address this issue,we propose the ST-ODBC model,a machine learning method for predicting ODBC rankings in road networks.The ST-ODBC model aims to predict the ODBC ranking of the next time slice based on the ODBC rankings of the previous k time slices.The ST-ODBC model mainly consists of a graph convolutional network(GCN),long short-term memory neural network(LSTM),and fully connected(FC)layer.The GCN is primarily used to capture spatial information,while the LSTM captures time series features.Through experimentation and analysis,the performance of the model is evaluated,demonstrating the effectiveness of the ST-ODBC model in predicting ODBC rankings in road networks.In addition,we provide a comprehensive review of the literature on betweenness centrality and its variants,as well as machine learning-based vertex ranking.