Research On Commuting Travel Mode Mining Of Taxi Traffic Zone Using GPS Trajectory Data

With the acceleration of urbanization,the spatial separation of urban residents’ work and residence has become more and more obvious,and the commuting pattern between urban residents’ work and residence has become more complex.It is convenient and common for city dwellers to commute by taking taxis between their homes and workplaces.At present,few scholars use taxi GPS data to explore commuting travel patterns.In fact,mining massive taxi GPS data and identifying taxi commuting patterns between regions can provide effective support for the transportation department to design reasonable public transportation services(such as customized buses).It has very important practical significance for improving the sharing rate of urban public transport and alleviating urban traffic congestion.This thesis takes the main urban area of Chongqing as the research scope,based on the GPS trajectory data of taxis in Chongqing,and uses data mining technology to carry out research on the identification of inter-regional taxi commuter traffic and urban hotspot route mining,and verify the effectiveness of the method by experiments.The main research contents are as follows:(1)Aiming at the problems of low efficiency and low recognition accuracy of taxi regional commuting identification methods,a taxi commuting traffic area identification method based on improved K-means algorithm is proposed.The method includes dividing traffic zones,generating flow transfer matrix between traffic zones,and identifying pairs of commuter traffic zones.Based on the existing division methods of traffic zone,a bottom-up division method based on fine-grained cell is proposed.Aiming at the problem of road OD allocation,a frequency-based method is proposed to allocate OD on roads to traffic zones.In the identification model,we take the peak-hour traffic flow and its dispersion coefficient as input features to mine the commuting traffic zone pairs.Finally,based on the real Chongqing taxi data set,the proposed method is verified.We identified 52 pairs of commuter traffic zones with significant commuting relationship,and analyzed the spatial and temporal characteristics of taxi commuting between traffic zones.The results verify the effectiveness of the method.(2)Aiming at the problems of incomplete consideration of factors and low accuracy of similarity measurement in the existing similarity measurement methods of trajectory clustering algorithms,this paper fully considers the temporal semantic characteristics and spatial semantic characteristics of GPS trajectories,and proposes a density core-based TP-DBSCAN(Trajectory Partition-DBSCAN)trajectory clustering algorithm.The algorithm mainly includes three parts: Trajectory segmentation based on RDP algorithm,spatiotemporal similarity measurement of GPS trajectory and clustering based on density core.Finally,based on the real Chongqing taxi data,the TP-DBSCAN algorithm is compared with the traditional TRACLUS algorithm and DBSCAN algorithm.The experimental results show that the TP-DBSCAN algorithm is superior to the compared methods in terms of clustering accuracy.