Urban Rail Transit Passenger Travel Behavior Analysis Methods Based On Cellular Data

Urban rail transit has many advantages like long distance,large capacity,comfortable,safety and efficient.It is backbone of urban comprehensive transportation system.Developing rail transit and improving rate of rail transportation will benefit to solve the urban traffic congestion.Collecting and analyzing traffic information by ITS technology is helpful for traffic operation management and traveler guidance,which will improve operation efficiency and service quality of rail transit.As a kind of big data,mobile cellular data has been widely used in traffic status analysis and individual or group travel behavior analysis.This thesis mainly focuses on travel behavior analysis of rail transit passengers and rail transit traffic status analysis for passenger travel guidance.The methodology architecture about individual travel activity analysis and passenger group travel behavior analysis are proposed.Analysis methods in current research generalized and divided into four kinds including matching,filtration,judgment,count.Then these kinds of methods are applied to progresses like geographic coordinate system and station ID matching,data pretreatment,historical data analysis,real time data analysis.Every mobile phone signaling events record is matched to certain geographic coordinate or station based on open source data and of mobile communication base station.The numbering plan is proposed to identify the station and line of rail transit.A space time constraints based geographic coordinate repair method is proposed to repair the signaling events record without geographic coordinate.The possible signal coverage area is positioned by using accurate positioning base station as datum point and passenger's upper mobility limit during the time passing three base stations as space time constraints.Calculations show that the error of corrected result is between 1 and 2 kilometers.This accuracy is adjacent to cell of origin positioning method itself and the repairing method is meaningful for high load base station without geographic coordinate data.The data pretreatment is proposed.Noisy data such as invalid redundant data,ping-pong data,drifting data is caused by the instability of mobile communication system.The data pretreatment aims to reduce the noisy data which will cause mistake or make no sense in next identification progress as more as possible.Also data pretreatment should try to reserve more data which can provide certain valuable information for next identification progress.According to this principle,the identification rule,identification technique process,threshold value of noisy data is proposed.Calculations show that,according to the threshold value infer from trial calculation and signal coverage feature,11 percents data is filtered as noisy data,which makes next identification progress easier.A space time constraints based passengers' activity trajectory identification method is proposed.Every passenger' s entrance station,exit station,transfer station,origination,destination can be extract from personal day signaling record sequence matched to geographic coordinate and rail transit station data.The personal activity trajectory identification progress takes the complexity of passengers' activities into consideration.Exit station identification uses proximity principle to find the line and station where passengers leave rail transit system,which reduces the dependence on survey on ground.Transfer station identification is divided into Signaling record list,station list,line list,transfer station validation and transfer station completion.This transfer station identification method can use not only normal position update events but also other random events.Based on passengers' activity trajectory identification result,every trip in one line can be matched to one train according to train schedule.The whole rail transit trip of one passenger can be represented.According to the station,line,train number,time information in personal activity trajectory,traffic status of line and station including inbound,outbound and transfer passenger volume,walking time out of train,passenger quantity in certain train or station space at certain moment,station service area.Inbound and outbound passengers,volume is validated by AFC based passengers' volume result.The walking time of transferring is validated by field survey result.Calculations show that the error is acceptable and the relevant analysis progress is feasible.The analysis method for real time data is also proposed.Real time data is different from historical data.In the analysis moment,a complete whole day signaling list has not created and the current signaling record cannot be analyzed based on future data.Inside rail transit system,signaling record is sparse except normal location update event,so different passenger characteristic parameters have different sampling characteristics.For high sampling rate parameters,a directly judging method is proposed.This method contains passengers' current state identification,historical state amendment.Directly judging method can extract real time volume of inbound,outbound and transfer passenger,number of passengers inside certain location area,walking time from train to ground outside the station.For low sampling rate parameters,correlation between passenger characteristic parameters and different signaling characteristic parameters is analyzed based on big data idea.Neural network is used to calculate the number of passengers in certain space.Based on correlation analysis,abnormal passenger flow warning method is proposed.Both passengers gathering inside rail transit system and inbound passengers gathering forward to certain station can be predicted.Based on the analysis methods of mobile phone signaling,the rail transit network of one megacity is chosen as case study.The results of historical data and real time data are calculated,validated and evaluated.Different results can be used in different scenes of rail transit organization and guidance.