Trip Chain Identification And Transfer Modeling Of Multi-Modal Public Transport Based On Multi-Source Data

Transport location data is important basic data for travel demand analysis,which can provide critical data support for managers to optimize transport services.The complete trip of a traveler can be represented by a trip chain,including travel mode,travel time,and location.The existing travel data from the transportation department is not available for the complete trip chain,including walking.The collection of Internet location data offers the possibility to carve out a whole trip chain.Current research on trip chains is based on GPS data and travel survey data mainly to analyze travel characteristics and build various algorithms such as random forests and decision trees to identify travel modes and travel purposes.However,existing studies have not yet verified the applicability of these models on Internet location data.By designing a self-collection scheme for mobile-based trip chain data,both Internet location data and travel mode data were collected.A trip chain identification model based on Internet location data is established,and the accuracy of the trip chain identification model is improved in terms of both data fusion and algorithm optimization.Based on travel feature extraction and trip chain modeling,a public transportation network based on metro and bus is studied,and a metro-bus transfer flow model based on public transportation network features is established.Then,the metro feeder mode is further extended to walking and bicycling.A more comprehensive multi-modal feeder travel model for metro stations is established,and the influence of land use and network characteristics of different areas around metro stations on metro feeder flow is analyzed.The specific research work is as follows.1.Study which trip data and travel features can be used to identify the trip chain.Transport data from transportation departments have travel mode markers but cannot cover the complete travel process,and Internet location data can cover the complete travel process but no travel mode markers.To verify the usability of the trip chain model on Internet location data,a mobile-based trip chain data collection scheme is designed to collect both Internet location data and travel mode data,and a travel feature importance selection model is established to provide effective explanatory variables for trip chain modeling based on Internet location data.2.A method to improve the accuracy of trip chain identification models is investigated.By building a multiple binary classification model based on data fusion,different sets of travel features are built for different travel modes,and then data from transportation departments are integrated into Internet location data in steps.The prediction accuracies of multiple binary classifications for walking,metro,bicycle,bus,and car are 0.839,0.809,0.735,0.799,and 0.799,respectively.Compared with the multiple classification models built based on a unified feature set,the accuracy of trip chain identification is improved from both data source and algorithm levels.3.Focusing on the public transport network based on trip chain analysis,the transfer model between metro and bus based on public transportation network features is studied.In addition to the transfer time,the number of stations,and the number of bus lines,the newly defined index "transfer accessibility" is proposed to characterize the combined effects of travel radius,the convenience of station transfers,and the public transportation service to which the station is connected.The method of calculating the transfer accessibility based on the gravitational model is proposed,which is proportional to the number of POI(Point of Interest)in the area accessible by public transportation and inversely proportional to the travel distance.All four features have a significant impact on the transfer flow,and the "transfer accessibility" improves the model accuracy from 0.6032 to 0.6935.4.A more comprehensive multi-modal metro connection prediction model is studied that extends metro connections to bus,walking,and biking.Identify the core circles and adjacent circles of the Voronoi diagram where the metro station is located,calculate the number of POI in different circles of the Voronoi diagram where the metro station is located,and use the metrics of the sub-circles of the Tyson polygon to characterize the different transfer attraction ranges of bus,bicycle,and walking.The relationship between the multi-modal metro transfer flow and the land use and network characteristics of the area around the metro is studied to break through the analysis of the land use area with the radius of the metro station and the walk-based transfer trips and to improve the accuracy of the metro transfer flow.Through trip chain data modeling and transfer network modeling analysis,this study provides theoretical research and practical analysis for trip chain identification and multi-modal public transportation transfer modeling based on multi-source data and also helps transportation planners to understand traveler behavior better and build a high-quality and efficient urban transportation management system.