Regional Accessibility Research Under Peak Hour Based On Clustering Of Taxi GPS Data

In traffic planning and research area,accessibility,an indicator evaluating the convenience of people or goods transportation,plays an important role in traffic policy research and transportation planning.However,accessibility has different definition in different research fields and its calculation method remains uncertain.It is also difficult to acquire the necessary data.As a consequence,few people actually use the approach of accessibility to guide or evaluate the transportation plans.Nowadays,the research in traffic big data matures and the relevant research about accessibility has begun.This dissertation analyzes the concepts and methods based on the review of relevant papers and proposes a method evaluating the regional accessibility based on clustering analysis of taxi GPS data.Firstly,this thesis discusses the the concept of accessibility and the reasons why it is difficult to define.The definition of the accessibility of this article is delivered afterwards.After determining the concept of accessibility and measurement methods,the regional accessibility study is divided into three sub-problems: network cost calculation,hotspots clustering and accessibility calculation.Clustering analysis of GPS data are proposed to determine the parameters needed for accessibility calculations.In the data cleaning stage,the paper proposes a batch data cleaning,correction and index calculation method,and applies this method to complete the data pre-processing procedure.After data pre-processing,the paper adopts an improved density clustering algorithm,and performs spatial clustering analysis on the points of arriving and departing taxis in Xiaozhai District,so as to determine the clustering object of accessibility calculation.After matching GPS data to road sections,the average travel speed and stop point counts are aggregated on road sections.Based on the statistics,K-Means and its improved methods are used to cluster road conditions to determine different condition types and travel cost.Finally,on the basis of hotspots clustering and network cost,the spatial impedance model,contour measure and cumulative opportunity model are used to calculate and evaluate the accessibility of Xiaozhai area.The calculation results show that in Xiaozhai area,the accessibility of hotspots clusters changes greatly during peak hours and nighttime;Xiaozhai's accessibility to hotspots is mostly within 15 min to 30 min;the contours feature the road network structure.