Cause Analysis Of Traffic Congestion:An Integrated Computational Graph And Multi-level Estimation Approach Based On Multi-source Data

In recent years,it has become increasingly apparent that urban traffic network managers are aiming to alleviate traffic congestion and improve the quality of service for congested streets.Traffic planners and managers are responsible for monitoring and analyzing the observed root sources of traffic congestion,especially under the emerging big data environment.This paper focuses on traffic congestion analysis problem,and proposes a multi-level computational graph framework that integrates a demand estimation model and a volume-delay function to capture and analyze the sources of traffic congestion from both the demand and supply perspectives.First,based on the traffic four-stage method,we present and analyze the influencing factors of traffic congestion from different levels,such as trip generation,trip distribution,route choice and road traffic condition.Second,we present a multi-level combined estimation model to simultaneously estimate different layers of variables from both demand and supply sides.Based on a customized computational graph framework,the combined model is reformulated as a multi-layer computational graph model and then is solved by the back propagation algorithm using heterogeneous data sources layer by layer.In the end,a series of case studies based on Sioux Falls network and Beijing Jingtong express road are implemented to demonstrate the effectiveness and applicability of the proposed methodology.The main research contribution of this paper is as follows:(1)we describe the relevant factors affecting urban traffic congestion by qualitative and quantitative analysis from multiple layers such as trip generation,trip distribution,travel route,link volume and link travel time;(2)we clarify the relevance of demand,supply and traffic congestion state using an integration demand estimation model and a customized flow-delay function;(3)we demonstrate the rationality and feasibility of the proposed model based on the Sioux Falls network case study.In the Jingtong Expressway Road case study,we capture demand sources of the congested road,analyze the demand characteristics under traffic congestion,and summarize the main causes of traffic congestion.