| Taxis attract a certain passenger flow because of their convenient,flexible advantages.As the urbanization rate increases,urban population density increases,urban traffic pressure is getting larger,and the problems accumulated by the taxi industry have gradually been highlighted.During peak travel periods and crowded areas,people need to line up to take taxis.During low travel periods,taxis run empty to find customers,wasting resources.The key to solving these problems is to master the rules and patterns of taxi travel and find the factors affecting the relationship between taxi supply and demand.The large number,large scale and strong mobility of taxis make it difficult to effectively understand taxi travel as a whole.Taxi trajectory data records the position and time of the vehicle,and contains the spatio-temporal characteristics.The comprehensive application of track data mining method can extract the spatio-temporal characteristics of the taxi trajectory data and reveal the nature of their occurrence.The work presented in this paper focuses on several aspects of the following: first,descriptive statistical method is used to mine the the periodic pattern of Taxi trajectory data;the hierarchical density-based spatial clustering of applications with noise(HDBSCAN)is used to extract hot spots of of trajectory data,and explore the the frequent pattern;second,POI data is used to quantify the urban built environment as independent variables,the taxi trajectory as dependent variable,the approach adopted is called ordinary least squares(OLS),geographically weighted regression(GWR),multiscale geographically weighted regression(MGWR)to identify the main factors of the taxi travel,check the spatial heterogeneity of the influencing factors.From these experiments one can conclude that:(1)In terms of research methods: HDBSCAN defines mutual reachability distance and cluster stability,performs significantly better and more robust than density-based spatial clustering of application with noise(DBSCAN)and hierarchical clustering.MGWR obtains the best model fit and explanatory power,OLS is the lowest.GWR assumes that all of the processes being modeled operate at the same spatial scale.MGWR not only is superior in replicating parameter surfaces with different levels of spatial heterogeneity but provides valuable information on the scale at which different processes operate.(2)In the spatio-temporal characteristics of taxi travel: overall within one month,the amount of taxi gradually increases with the date,repeating the cycle in a week;8:00 to10:00 in the morning,13:00 to 15:00 at noon,from 17:00 to 19: 00,are the peak period of travel,and the length of the single travel is about 15 minutes;workdays and rest days have significant differences;among the different time cycles,the space distribution of the hot spots is basically consistent,gradually increased from the Chengdu 3nd Ring Rd to the Chengdu 1nd Ring Rd,and divided into typical land use areas,such as business cluster,residential land,external transportation,scenic spots,mixed land;the change of hot spots in different time periods is most obvious in the business cluster;by comparing the distribution of starting point and end point,the clustering number of the end point is more than that of the starting point,and the clustering range is more concentrated and the coverage is smaller.(3)In terms of influencing factors: the urban built environment has a strong effect on taxi travel;catering service,corporate business,shopping service,land-use mix,recreation and entertainment,bus station,road density,population density affect the number of trips;population density has both positive impact and negative impact;shopping service is affected by the scale of spatial division;POI data can effectively quantify the urban built environment and guarantee the comprehensiveness of impact factors.In short,this paper establishes the multi-character recognition,multi-character expression and multi-scale modeling of the taxi trajectory. |
PDF Full Text Request