Analyzing Spatiotemporal Association Of Traffic Congestion Based On Empirical Data In Tianjin

Urban transport system is the basic guarantee of urban life.lt can promote economic and society development.However,the rapid growth of people and vehicle ownership has caused the conflicts between transport demands and supplies,leading to traffic congestion,pollution,and safety problems.Preventing and relieving methods based on the analysis of spatiotemporal association of traffic congestion can reduce its frequency and influence.Most of existing researches focus on spatial or temporal assoiciation separately,ignoring the association between space and time.Therefore,this thesis proposed the recognition method for traffic congestion,and the analysis method of spatiotemporal association for congested area based on the data ofroad network.Firstly,this thesis recognized traffic statuses.Fuzzy Clustering Algorithm(FCM)was selected for traffic status recognition according the uncertainty and fuzziness of traffic status definitions.Kmeans++algorithm was used for determining the initial cluster centers,in order to overcome the sensitivity of FCM to initial cluster centers,which improved clustering effect.Then,this thesis analyzed the spatial or temporal distribution characteristics for different traffic statuses,which came to the conclusion that the amount of congested areas is highest around 8am and 17pm for weekdays,and 10am and 17pm for nonworking days;and most eongested areas inelude metro stations,schools,flyovers and regions nearby.Secondly,this thesis studied on the association between congested areas.Spatiotemporal association rules,including rules within transaction and inter-transaction rules,were disigned.Rules within transaction were used for analyzing the correlation between congested areas at the same time.Inter-transaction rules were used for analyzing the association between congested areas at different times,aiming to find spatiotemporal evolution patterms of traffic congestion.Then,in order to solve the algorithm problems of repeated scanning database,calculating support and large candidate frequent sets,three improvement strategies were proposed accordingly from aspects of algorithm logic and data structure.A case study was carried out and result showed that the computational efficiency of optimized algorithm was improved.Finally,this thesis applied the optimized algorithm of spatiotemporal association rules to analysis spatiotemporal association of traffic congestion in Tianjin.FCM was used to recognize congested areas,which can form the transactions.By applying the rules within transaction to transactions,the correlation between congested areas at peak times can be received.And inter-transaction rules were applied to analyzed spatiotemporal evolution patterns of traffic congestion at peak times.Results showed that congested areas are related both spatially and temporally.Congested areas are not only related to each other at the same time,but also related at different times.By mining the spatiotemporal association of congestion can deepen the understanding of the generation and evolution pattern of congestion.Thereby it can provide scientific support for congestion management and control strategies,and improving road operation and congestion management level.