REAL-TIME BUS TRAVEL TIME VARIABILITY AND RELIABILITY ANALYSIS: AGTFS-RT APPROACH

Abstract

In publictransport,traveltimereliabilitymeasurestheextentofunexpecteddelays on transitroutes.Itisoneofthecrucialfactorsaffectingpassengers’behaviour,including choicesofroute,departuretime,andmode.Thisstudyfocusesonenhancingpublictrans- portefficiencybyestimatingtraveltimevariabilityontransitsegments.Itismeasured using standarddeviationandthecoefficientofvariationoftraveltime.Additionally,a bus arrivaltimepredictionmodelisdevelopedusingvariousmachinelearningandartifi- cial neuralnetworkmodels.Fortheestimationoftraveltimevariabilityontransitroutes and segmentsforthearrivaltimepredictionmodel,thecurrentstudyutilizedonemonth (May2023)ofreal-timeGTFSdatafromDelhi,India.Overallanalysisshowsthattravel time variabilityishigheronThursdayandloweronMondayamongallroutes.Theresults indicate thatboththedayoftheweekandthetimeofdaysignificantlyinfluencetravel time variabilityontheroutelevel.Further,analysisshowsthatthemaximumtraveltime variabilityonthesegmentsisobservedduringoff-peakhourscomparedtoothertimes of theday.Furthermore,theresultsfromthepredictionmodelsshowthattheextreme gradientboostingmodeloutperformstheothermodels.Thepredictionmodelsprovide arrivaltimesatthetransitstops,helpingtransitpassengersplantheirtripsandreduce their waitingtimeattransitstopsbyplanningahead.Fromtheperspectiveoftransit operators,thevariabilityanalysiscanbeusedtoimprovetheperformanceoftransit services ondifferentsegmentsoftheroutes.