Graph Database Solutions for Managing and Analysing Spatio-Temporal Data in Transportation Networks

Abstract

In thisthesis,weadaptandextendthetheoryoftheexistingpropertygraphdatabases to modeltransportationnetworks,wherenodesandedgescancontaintemporalproperties that aretimeseries,resultinginPropertyGraphswithTimeSeries.Weproposealanguage based onregularpropertygraphlogicforqueryingthesePropertyGraphswithTimeSeries, in whichtime-seriesandmeasurementpatternsmaybecombinedwithgraphpatternsto describe,retrieveandanalysereal-lifesituations.Further,wedemonstratethemodeland language inpracticebyleveragingtheGraphPatternMatchingLanguageasabridgebetween theory andpractice.TheproposedgraphmodelisimplementedinNeo4jwithagraph- embeddedstructureforthetimeseries.ByleveragingNeo4js’queryengine,werealise the querylanguagewithatranslationlayer.Usecasesexplorequestionsthathydrology researchersposeinthecontextofthe“InternetofWater”,electricalengineerscanposeon localdistributiongridsandrailwaymanagersconsiderindailyoperationmanagement.In addition totheanalyticalcapabilitiesprovidedbythequerylanguageandmathematical expression implementedinexistingdatabasesystems,amoreOLAP-likeaggregationand exploration mechanismisresearched.Withadefinedhierarchystructure,aggregationtrees can bebuiltontopoftheoriginaldata,allowinguserstoexploredataindifferentspaceand time levelcombinations. This way,wewanttoprovideaunifieddatamanagementsolution,includingstorageand data analysiscapabilities,thatcanbeusedbydomainexpertswithgeneraldatabaseknowl- edge. Indoingso,reducingtheset-uptimeandmaintenanceneededinresearchprojects using sensornetworksthattrytosolvetransportationnetwork-relatedresearchquestions.