Motivation

DataVisualizations

RendezView Interface

research

edinburghdata.intensive

Acknowledgments

Theprimaryvisualizationisa,implementedusingThree.js5 and

GeoJSON2 data.GeospatialdataisrepresentedintheX-Y(red/blue)dimensions,whiletemporaldataisrepresentedintheZ(green)dimension.Eachcubemappedontoprepresentsamatchingrowfromthedatabase.

The showsthefrequencyofkeywordsandhashtagsusedinconjunctionwiththesearchedonkeyword.Itupdatestoincludetheaggregateofallwordfrequencyinformationwhenmultipleboxesareselected.ThiswasimplementedusingD3.js1.

The consistsofnodesandlinks,wherenodesarekeywordsandlinksaregeospatial-temporalintersectionsbetweenkeywords.Thewidthofthelinksshowstheflowquantity,whichisthemeasureoftheintersectionoftheboxesinthe3Dmap.ThisisbasedonaD3Sankeyplugin.

Todevelopaweb-basedinterfaceforvisualdatamining3 toextendthefunctionalityofthepreviousSophy4 framework:• Implementmultiple,interactivedata

visualizations:geo-spatial,time,&topicdata.• Improveuserinteractioncoupledwithdata

miningprocesses:datafiltering,selection,aggregation,etc.

RendezView:AnInteractiveVisualMiningToolforDiscerningFlockRelationships

inSocialMediaDataMelissaBica1 and Kyoung-SookKim2

1DepartmentofComputerScience,UniversityofColorado,Boulder,CO2ArtificialIntelligenceResearchCenter,AIST,Tsukuba,Japan

2

3

4

1

FutureImprovements

• Connectiontolivedatabase.• Sankeydiagramfunctionalityandinteractionwith3Dmap,includingoptiontoviewlinkwidthsaccordingtovariousaggregationtypes.

• Front-endperformanceimprovements.

Figure1. TheabovescreenshotshowstheUIlayout,design,andfunctionalitiesoftheRendezView framework.(1) Theuserinputstheiroptionsforfiltering thedatatodisplay.(2) Databaseentriesthatmatchtheselectedfiltersarerepresentedonthe3Dmapascubes.Eachcube’sposition,dimensions,andcolorcorrespondtogeospatial,temporal,and/ortopicmetadata.(3) Additionalvisualizationsappearwhenboxesareclickedon.Theseincludeawordcloud,a2Dmaptomoreclearlyshowgeospatialdimensions,andlabelsdisplayingthetimerangeoftheselecteddata.(4) ASankeydiagramshowsflockrelationships betweenkeywords.Theusercanclickonlinksbetweenkeywordstodisplaygeospatialtemporalintersectionsofthosekeywordsonthe3Dmap.

Summary

References

[1]D3:http://d3js.org/[2]GeoJSON:http://geojson.org/[3]Keim D.InformationVisualizationandVisualDataMining.2002.[4]KSKim,HOgawa,ANakamura,IKojima.Sophy:aMorphologicalFrameworkforStructuringGeo-referencedSocialMedia.2014.[5]ThreeJS:http://threejs.org/

TheauthorsofthisprojectthankthePartnershipinInternationalResearchandEducation(PIRE)program,theOpenScienceDataCloud(OSDC),theNationalInstituteforAdvancedIndustrialScienceandTechnology(AIST),andtheNSFfortheirsupport.Additionally,wethankDr.BobGrossman,Dr.MariaPatterson,andDr.JasonHaga fortheirguidance.

RendezView isaninteractive datavisualizationframeworktoshowflockpatterns andrelationships incomplexdata,particularlydatafromsocialmediaplatformssuchasTwitter.“Flock”referstotheclusteringofsimilardata.Themultipleinteractivevisualizationsallowuserstoassessmanylevelsofrelationshipsintheirdataandeasilydetectpatternsthatwouldnotbeevidentinadatabasealone.ItisbuiltinHTML5,CSS,andJavaScript.

InteractiveVisualDataMining

Figure2. RendezView isatoolforvisualdatamining:theprocessofdetectingpatternswithinbigdatausingvisualizations.Thisdiagramshowshowdataisrepresentedandtheinteractionsbetweenthevisualizations.

UseCases

RendezView canbeusedtoinvestigatedifferentsocialphenomenaoverageographicregion,suchasinformationflowindisastersandothercrisisevents,patternsinworkandhiring,ortrendsinpoliticaldiscourse,byiterativelyusingresultsfromeachvisualizationtoinformsubsequentinteractionswiththedata.Asanexample,asocialscientistcancomparetheflockpatternofjobpostingsbetweentheeastandwestcoastsintheUSbysearchingonthekeyword“work.”Theresearcherlooksatwhichkeywordshaveasimilarpattern,suchas“apply”and“hire.”TheSankeydiagraminformsthenextkeywordsearchbasedonstrongspatiotemporalco-occurrencefrequencywiththeoriginal.