Modeling of Transport Demand Analyzing, Calculating, and Forecasting Transport Demand V.A. Profillidis G.N. Botzoris Elsevier Radarweg29,POBox211,1000AEAmsterdam,Netherlands TheBoulevard,LangfordLane,Kidlington,OxfordOX51GB,UnitedKingdom 50HampshireStreet,5thFloor,Cambridge,MA02139,UnitedStates Copyright(cid:1)2019ElsevierInc.Allrightsreserved. Nopartofthispublicationmaybereproducedortransmittedinanyformorbyanymeans, electronicormechanical,includingphotocopying,recording,oranyinformationstorageand retrievalsystem,withoutpermissioninwritingfromthepublisher.Detailsonhowto seekpermission,furtherinformationaboutthePublisher’spermissionspoliciesandour arrangementswithorganizationssuchastheCopyrightClearanceCenterandtheCopyright LicensingAgency,canbefoundatourwebsite:www.elsevier.com/permissions. Thisbookandtheindividualcontributionscontainedinitareprotectedundercopyrightby thePublisher(otherthanasmaybenotedherein). Notices Knowledgeandbestpracticeinthisfieldareconstantlychanging.Asnewresearchand experiencebroadenourunderstanding,changesinresearchmethods,professionalpractices, ormedicaltreatmentmaybecomenecessary. Practitionersandresearchersmustalwaysrelyontheirownexperienceandknowledgein evaluatingandusinganyinformation,methods,compounds,orexperimentsdescribed herein.Inusingsuchinformationormethodstheyshouldbemindfuloftheirownsafetyand thesafetyofothers,includingpartiesforwhomtheyhaveaprofessionalresponsibility. Tothefullestextentofthelaw,neitherthePublishernortheauthors,contributors,oreditors, assumeanyliabilityforanyinjuryand/ordamagetopersonsorpropertyasamatterof productsliability,negligenceorotherwise,orfromanyuseoroperationofanymethods, products,instructions,orideascontainedinthematerialherein. LibraryofCongressCataloging-in-PublicationData AcatalogrecordforthisbookisavailablefromtheLibraryofCongress BritishLibraryCataloguing-in-PublicationData AcataloguerecordforthisbookisavailablefromtheBritishLibrary ISBN:978-0-12-811513-8 ForinformationonallElsevierpublicationsvisitour websiteathttps://www.elsevier.com/books-and-journals Publisher:JoeHayton AcquisitionEditor:BrianRomer EditorialProjectManager:AndraeAkeh ProductionProjectManager:PaulPrasad CoverDesigner:MarkRogers TypesetbyTNQTechnologies To my wife Areti and my son Aristeidis V.A. Profillidis To Nikiforos, Zoe, and Vassiliki G.N. Botzoris Preface Forecastingthefuturehasalwaysbeenachallengeforhumanityinitseffortto safeguard achievements andbetterorganize future endeavors. Thefuture is an immense set of eventual plausible alternatives and generates hopes and fears. However,fromtheveryearlystepsofrationalthinking,manhasrealizedthatthe futureprovidesearlysignsofwhatwillhappen.Itrestsuponmantodepictand record these early signs, and based upon them, to understand and explore plausible future evolutions, and forecast as accurately as possible what will probablyhappen.Thus,thecoreofanyscienceistoanalyzedataandtheevents ofthepastinordertoforecastandcalculatethefuturevaluesofaphenomenon under study. Sciences related to inert materials (like physics and engineering) cancompileknowledgeinsomekindofuniversallaws,whicharevalidevery- whereacrosstheworldandforeverymomentinthefuturewithahighdegreeof certaintyandobjectivity.Thisisnotthecasewithsciencesrelatedtoeconomic andsocialphenomena,whichcanhardlyleadtouniversallawsbutoftenforesee futureevolutionswithadegreeofuncertaintyandsubjectivity,andfurthermore onlyforasmallrangeofproblems. Theaimoftransportistocarryapersonoragoodfromoneplacetoanother by using a specific transport mode (car, train, airplane, ship) which runs on appropriateinfrastructure(road,railwaytrack,airport,port).Transportpermits people to work at a distance of many kilometers from home, to buy and sell goods from remote areas thousands of kilometers from the productionor con- sumptionarea,andtovisitotherpeopleanddiscoverandexploreotherareasand civilizations even at the most distanced areas of the globe. Transport and (nowadays)theInternetgivepracticallyinfinitepossibilitiesforanyonetoextend andbroadenhisdomainofactivitiesandinterests;thisnewrealityhasinmany waystransformedthewholeplanetintoanewkindofglobalvillage. Transportdemandaimsatcalculatingtheamountoftransportthatpeoplewill chooseanduseforaspecificpriceorothercharacteristics(suchastraveltime, frequency,punctuality,etc.)ofatransportmode.Itmayconcernbothtransport modes and transport infrastructure. Transport demand models are simplified constructs that attempt to describe, explain, correlate, and forecast transport demand as accurately as possiblewithin a rational framework of assumptions. Transportdemandistheoutcomeoftechnicalaspects(suchastraveltimesorthe technology used for the transport mode and for the infrastructure), economic aspects(suchastravelcosts,revenueofthecustomers),lifestyleaspectsofthe xxi xxii Preface societyorpreferencesofaspecificindividualforatransportmode,psycholog- icalaspects,andsoon.Thus,atransportdemandmodelshouldtakeintoaccount both objective aspects of demand (related to the technological and physical characteristicsofthetransportactivity)andsubjectiveones(relatedtoeconomic and social conditions). Transport demand models try to establish the most accuraterationalcorrelationbetweentransportdemandandthefactors(ofboth objectiveandsubjectivecharacter)thataffectit;thiscorrelationcanbeconducted atvariouslevelsofanalysisandsophistication,frompurelyqualitative,basedon moreorlesssubjectiveperceptionsandassessmentsmadebysomeindividuals,to quantitative ones, which are based on some form of mathematical relationship betweentransportdemandandoneormorefactorsaffectingit.Suchacorrelation shouldbeexplored,established,checked,andvalidatedbeforeitcanbeusedfor theforecastoffuturetransportdemand. Thebasiccomputationaltoolsforconstructingatransportdemandmodelare statisticsandcomputationalintelligence.Statisticspermitatransportforecaster toanalyzedataofthepast,toformulateanequationwhichdescribesaccurately theevolutionofthesedata,andthentousethisequationfortheforecastoffuture transportdemand.Theequationofatransportdemandmodelmaydependonly on time or (in addition to time) on other factors which also affect transport demand.Computationalintelligenceisusuallyemployedwhenmethodsbased on statistical techniques fail to accurately simulate a problem; computational intelligence techniques make use of the knowledge and methods of artificial intelligence,andparticularlyofneuralnetworkmethodsandoffuzzymethods. Transportdemandmodelingisarathernewbranchoftransportsciencewhich began to develop during the 1960s alongside the construction of new big infrastructure projects. Up until a few years ago, the basic tools for transport demand modeling were qualitative and simple statistical methods. However, evolutions in computer science and new ways of mathematical thinking have broadenedthefieldofmethodsandtechniquesfortheanalysisoftransportdata andtheforecastoftheirevolution.Thus,whenfacinganyproblemoftransport demand,aforecasterhasaccesstoaplethoraofmethods,bywhichheshouldbe neither impressed nor deterred. The central aim of the book is to analyze and presenttheavailablemethodsofforecast,theminimumtheoreticalbackground foreachoneofthem,theappropriateformulasandcomputersoftwaretobeused in each case, and applications to specific problems. However, the forecaster shouldnothaveanykindofprejudiceinrelationtothedegreeofsophistication of a method: no matter how mathematically complicated or simplistic, any method of forecast can be selected if it can afford forecasted values close to reality, provided that it is previously tested for its validity and accuracy. The knowledge included in the book comes from original and applied research conductedbytheauthorsaswellasbyotherspecialistsofthesectorworldwide. The science of transport demand modeling has thus far been developed empirically on the occasion of specific problems in need of a solution and is not founded on solid theoretical considerations and background. As a result, a Preface xxiii forecaster facing a specific problem typically comes across numerous case- studies, without a clear identification of assumptions, methods, and computa- tional techniques used. In this book we have put great effort into founding transportdemandmodelingonamoreaxiomaticapproach,byexplicitlygiving the assumptions for each theory and method and testing the formulas and computer software for their validity. We have tried to reduce the number of assumptions and conditionsofapplication ofa methodasmuchaspossible, in ordertoprovidetothesimulatedformulationoftherealproblemthemaximum number of degrees offreedom, so as to represent as accurately as possible the real problem under study. Transportdemandforecastsareaprerequisiteforalmostallactivitiesrelated totransport.Anydecisionrelatedtotheplanning,investment,andoperationof both the infrastructure and thevarious transport modes always needs the most accurate forecast of transport demand, and even more so nowadays in the competitivenationalandinternationaleconomicenvironment.Anyconstruction of a transport infrastructure or any operation of a transport services company whichisnotbasedonthemostaccurateforecastoffuturedemandrunstheriskto turnintoaneconomicadventureandoftenineconomicdisaster. Transportdemandforecastscanbeconductedbyspecialistscomingfroma scientifically diversified background: engineers, economists, mathematicians- statisticians, computer and data analysts, andso on.The bookaddresses to all these specialists, either working in the field or studying in schools of engi- neering, economics, transportation, applied statistics, computer science, and businessadministration.Abigdilemmafacedatthebeginningofthewritingof the book related to the minimum prerequisite mathematical, and particularly statistical, background that the reader ought to have in order to read and un- derstandin-depththecontentofthebook.Weoptedforaminimumbackground atthelevelofafirstyearstudentofengineeringoreconomics.Forthisreason,we provide (in Chapter 5 and wherever it is necessary) a short analysis of the requiredtermsandknowledgeaboutstatisticsandmathematics.Thus,thebook canbeeasilyreadbyanystudentstudyingintheaforementionedfields,butalso any professional working in the transport sector after decades from his gradu- ationandtheinevitableerosionofmathematicalknowledgeinhismind. The book draws knowledge from many scientific branches, such as eco- nomics,engineering,mathematics,statistics,informatics,andevenroboticsand psychology.Asthereadermaybeinterestedinaspecificpartofthebook,wedid nottakeforgrantedthathehasalreadycomprehendedwhatprecededhisspecific area of interest. For this reason, the reader is systematically referred towhich other part of the book the necessary knowledge for understanding his specific topicofinterestismadeavailable. Thefirststepinanyeffortoftransportmodeling,andbeforelookingforthe appropriatemethodtouse,liesinthattheforecasterunderstandsclearlywhich arethefactorsordrivingforcesthataffectthespecifictransportdemandprob- lem. All these factors and the related elasticities are analyzed in Chapter 1, xxiv Preface together with the analysis of whether rates of growth of transport are similar (coupling) or not (decoupling) to rates of growth of gross domestic product (GDP). As transport is a dynamicphenomenonevolving over time, the analysis of trends is a second prerequisite before modeling. Chapter 2 makes a profound analysis,overthepastfourdecades,ofpassengerandfreightdemandforair,rail, road,andseatransport.Theanalysisisconductedbothatworldlevelandatthe levelofmanycountries. Beforeselecting any modeling method,it is necessary fortheforecasterto have a panorama of thevarious methodsas well as of the successive steps for constructingamodel.InChapter3,theclassificationofmethodsintoqualitative (executive judgment, Delphi, scenario writing, questionnaire survey) and quantitative(trendprojection,time series,econometric,gravity)is formulated, andeachparticularmethodissuccinctlyanalyzedtogetherwithcomputational intelligence methods (artificial neural networks, fuzzy method). In addition to ordinary data, the application of “big data” is extensively analyzed. Utility theories, which can help to explain how an individual makes his choice for a specifictransportmode,arepresented. Thevarious qualitative methods are extensively presented and analyzed in Chapter 4 of the book along with the assumptions, characteristics, scientific background, and applications for transport demand problems. The executive judgmentmethodistheoutcomeoftheevaluationandassessmentofaspecific problembyspecialistsinaparticularsector.TheDelphimethodisbasedonthe judgmentofagroupofexpertsinsuccessiverounds;thedegreeofconsensuscan be evaluated by means of the Kendall’s coefficient of concordance. Scenario writing traces the pathways and describes the conditions of a new transport situation,byexploringtheeffectsofpast,present,and(likelyorunlikely)future events, and can be projective (from the present to the future) or perspective- visionary (from the future to the present). Survey methods(of both stated and revealed preference) can monitor, detect,and quantify, based on questionnaire responses,theattitudes,trends,andprospectsofcustomersofatransportservice and hence of transport demand. For all qualitative methods, a number of applicationsfortransportdemandproblemsareillustrated. Anymodelingmethodisbasedonanumberofstatisticalmethods,suchas simpleandmultipleregressionanalysis,whichpermitthecorrelationoftransport demandwiththefactorsaffectingit,andtheseareexplainedinChapter5ofthe book.Thevariousstatisticalmetrics,tests,andmethods,assuringthatregression analysisrepresentsaccuratelyatransportdemandproblem,areclarified:Pearson correlation coefficient, coefficient of determination, Student’s t-test, F-test, multicollinearity test (detecting eventual correlation between independent variables),testsrelatedtoresiduals,determinationofoutliers,existenceornotof serialcorrelationintheresiduals,heteroscedasticitytests,testsfortheevaluation oftheforecastingaccuracyandforecastingability.Alltheaboveareextensively putintopracticeinaspecificexampleofmultiplelinearregressionanalysis. Preface xxv Time series and trend projection, which are the two most commonly used quantitative methods for the forecast of transport demand, are analyzed in Chapter6ofthebook.Thevariouscomponentsofatimeseries(trend,seasonal, cyclical, random) are identified, and a methodology for a seasonally adjusted trend projection is presented. Assumptions, methodology, successive steps, statistical tests, and the appropriate software and specific case-studies for the trendprojectionmethodofforecastoffuturetransportdemandareallanalyzedin detail.Itisquiteusefultoconductanexpostassessmenttoverifyortoreconsider previousforecasts.Othersimplestatisticalmethods,suchasmovingaverageand exponential smoothing, are given. The various time series processes (white noise,randomwalk,autoregressive,movingaverage,(seasonalornonseasonal) autoregressivemovingaverage)areanalyzedandmethodstocheckstationarity of a time series are presented. The BoxeJenkins method, the most popular iterativeprocedureofmodelingatimeseries,isexplainedindetail,togetherwith anextensivecase-studyapplication. A causal correlation of transport demand to its generating forces (some of which are the independent variables) usually necessitates implementation of econometric models, which are the topic of Chapter 7 of the book. The suc- cessivesteps,theassumptions,andthenecessaryconditionsfortheconstruction of an econometric model are described together with how the independent variables, the functional form, and the statistical tests should be selected. The statisticaltestswhichcertifythevalidity,accuracy,andforecastingabilityofan econometric model are analyzed in detail. The appropriate econometric equa- tionsforair,rail,roadtransport(bothforpassengerandfreight),publictransport, androadsafetyaregiventogetherwiththestatisticaltests thattestifyfortheir validity.AnessentialpartofChapter7isdevotedtothefour-stepmethodforthe forecast of transport demand at urban level. The various models for the four successivestepsareanalyzed:tripgenerationdfirststep(growthfactormodels, cross-classification models, regression models), trip distributiondsecond step (constrained growth factor models, gravity models), choice of transport modedthirdstep(LogitandProbitmodels),tripassignmentdfourthstep(allor nothing model, user equilibrium model, system optimum model). The appro- priatemodelingoffreighttransportdemand(trendprojection,four-stepmodels, econometricmodels)isalsopresented. The last two chapters of the book deal with applications of computational intelligencemethods,whichtrytomakemachinesdothingsthatwouldrequire humanintelligence,thatis,tothinklikepeople,toactlikepeople,andtothink rationallyandreasonably.Chapter8dealswithartificialneuralnetwork(ANN) methods, which attempt to imitate the way biological neurons are operating (externalinput,activationornot,output).ANNcanbeusedforshort-andlong- term forecasts of transport demand and can consider nonlinearities, a great number of input (independent) variables, and a tremendously high number of data.Howanartificialneuronoperates,thedifferentformsoflearning,andthe variousalgorithmsofANNareexplained.However,ANNdoesnotprovideany xxvi Preface interpretationofthemechanismortheprocessofaproblem.Detailedanalytical examples of how ANN can be applied for transport problems are studied and applicationsofANNfortheforecastofair,rail,roadtransportdemand,forthe schedulingoffreighttransport,andforassessingmaintenanceneedsaregiven. Timeseries,econometric,andANNmethodsarebasedonthemeasurability ofaproblemandemployrealnumbers.Incontrast,afuzzynumberdoesnotrefer toonesinglevalue(asanordinaryrealnumber)buttoasetofplausiblenumerical valueswithinaspecifieddomainandaroundacentralvalue.Fuzzynumbersare suitableforrepresentingphenomena,liketransportdemand,withvariationsor oscillationsaroundanumericalvalue,andcancombinesubjectiveandobjective knowledgeaswellastakeintoaccountuncertainty,imprecision,ambiguity,series ofdatawithmissingvalues,linguisticvariables,andevennonmeasurabledata. The properties and mathematical description of fuzzy numbers are described. Detailedanalyticalexamplesoftheapplicationoffuzzymethodsinregression analysisandforeconometricmodelsarestudied.Applicationsoffuzzyregression analysisfortheforecastofair,rail,roadtransportdemand,roadsafety,logistics androutingoffreightvehicles,andtheoptimizationofairportsarepresented. SometechnicalassistancewasprovidedbyMr.AlexanderYiannopoulos,an Americanspecialistinphilosophy.Wewouldliketoexpressoursincerethanks tohim. The book was written parallel to our scientific, professional, and family obligations.Wethankourfamiliesandcollaboratorsfortheirunderstandingand patience. September2018 V.A.ProfillidisandG.N.Botzoris Chapter 1 Transport Demand and Factors Affecting It Chapter Outline 1.1 TheBasicDefinitionsRelatedto 1.6 HowTransportDemandAffects TransportDemand 1 thePlanningoftheTransport 1.2 HistoricalEvolutionofTransport System 19 ModesandDemand 7 1.7 TransportDemandand 1.3 HowTransportDemandAffects EconomicActivity:Couplingand theTransportSystem 8 Decoupling 20 1.4 TheOverwhelmingand 1.8 FactorsoftheInternaland RevolutionaryEffectsof ExternalEnvironmentof NewTechnologies 12 TransportandEffectsonDemand 34 1.5 EvolutionOverTimeofthe 1.9 TransportDemandand PrincipalDriversofTransport Elasticities 44 Demand 13 1.1 THE BASIC DEFINITIONS RELATED TO TRANSPORT DEMAND 1.1.1 Transport and Human Life Transport, along with food, health, energy, and home, is an essential and crucial activity of human life. Without transport the world of every human beingwouldbelimitedtoafewdozenkilometersaroundtheareaofhisorher settlement. It is with transportthat mankind has tried from the prehistoric era to broaden its horizons and the scale of its activities, to look for better con- ditions for survival and prosperity. Only in the case of very primitive human life,wasmanliving,producing,anddevelopinginanarrowgeographicalarea. However, from the time of the first forms of human social organization, humans have realized the essential importance of meeting other people, exchanging experiences, coordinating their lives and strategies, discovering ModelingofTransportDemand.https://doi.org/10.1016/B978-0-12-811513-8.00001-7 Copyright©2019ElsevierInc.Allrightsreserved. 1