Research On The Influence Mechanism Of Urban Land Use On Car-Hailing Ridership Based On The Division Of Traffic Zones

With the acceleration of the urbanization process in China and the increase of car ownership,urban congestion caused by the imbalance between traffic supply and demand are becoming increasingly serious.Simply relying on the expansion of supply cannot solve the contradiction from the root,which is contrary to the concept of sustainable development.Only from the perspective of urban planning can the problem be solved by promoting the coordinated development of traffic and land use.However,most previous studies are based on traditional planning data,which are not timely updated,time-consuming and laborious in data acquisition.Besides,there are many studies on bus and rail ridership,and lack of in-depth exploration on floating cars and online car-hailing in the sharing economy.In recent years,online car-hailing as a new mode of transportation,has been welcomed by both drivers and passengers because of its convenience.Meanwhile,the resource sharing between travel service companies and the public,the deep integration of big data and mobile technology also provides an open channel and new research means for the acquisition and analysis of ride-hailing data.Based on the large-scale Di Di ride-hailing data and POI data,this paper proposes a two-level partitioning method for traffic zone division and takes the divided traffic zones as the basic unit.On this basis,the influence mechanism of land use on ride-hailing trips is deeply explored and the ride-hailing trips in traffic zones are predicted.The main research contents of the study are as follows:(1)The core area of Beijing fifth ring is determined as the study area of this paper.The study introduces the data structure of Di Di car-hailing ridership,cleans and filters its coordinates,analyzes the characteristics of spatiotemporal distribution of car-hailing ridership,and identifies the differences between weekdays and weekends in travel pattern and hotspot area.POI data based on Internet is also introduced and reclassified into six categories after being deduplicated,filtered and merged.The characteristics of the spatial distribution of different twelve land use types are analyzed,which lays a foundation for further study.(2)Based on the two-level partition theory,the traffic zone division is established from the big data perspective.A grid model is established.Using Di Di data that can reflect travel patterns and POI data that can represent land use features,the study area is firstly divided into several traffic medium zones with high internal connection and weak external connection through the community detection algorithm based on module degree.Under the guidance of the principle of homogeneity of traffic zone division,then,for each traffic medium zone,clustering method is used to cluster the grids with similar travel patterns and land use attributes into several traffic small zones.The study area is divided into 402 traffic small zones in two steps.(3)Taking the divided traffic zones as the basic unit,the study analyzes the influence of land use on travel pattern.The hourly distribution curves of ride-hailing travel attraction and generation density are drawn,and the difference between them is strongly correlated with land use properties.The net inflow density matrix(DPDM)is constructed to represent the travel pattern of the traffic zone.After the net inflow density of ride-hailing in 402 traffic zones is identified by the time series clustering algorithm,the correlation between land use and ride-hailing travel pattern is quantitatively analyzed based on the two methods of land use function index and correspondence analysis.Finally,on the basis of clarifying the corresponding relationship between the ride-hailing travel pattern and the land use,by analyzing the geographic location and spatial distribution of the traffic zones,the urban space structure was semantically calibrated.(4)Taking the divided traffic zones as the basic unit,the study analyzes the influence of land use on car-hailing ridership through the ordinary least square regression model(OLS)and geographically and temporally weighted regression model(GTWR).Finally,the density of land used for residential,industrial,commercial service,public facilities,green space,transportation facilities,and the density of subway station and bus station representing the characteristics of public transportation land were determined as the input independent variables of the regression model.The traffic generation and attraction of working and non-working days are four dependent variables,and the average effect of each influencing factor is preliminarily analyzed by the ordinary least square regression model,and the spatio-temporal heterogeneity of the impact of land use factors on online ride-hailing trips is deeply explored by the geographically and temporally weighted regression model.Finally,by comparing the relevant diagnostic indicators of the two models,it is found that the GTWR model has a better fitting degree for the spatial and temporal data,and has better prediction accuracy with the random distribution of prediction error terms.