Interactive Poster: Visual Mining of Business Process Data
MingC.Hao,DanielA. UmeshwarDayal, Schneidewind Hewlett ResearchLaboratories
1.
analyses,partialmatchingtechniques,aswellascluster,andclassificationanalysis.
Business 'datais largeand mast a triple-parameter symmetric circular graph layout to
to be directly Usually the represent the source, and destination
operationsconsistofmany and and attributes
every data may take a path the a process flow matrix to link multiple circulargraphs
In weshowa processschema. together showthe important ofthebusiness
this
processisaverysimple realistic processoperations.Inthecreditcard analysisexampleshowninFigure we business atleast10times
mayuseaclusteringalgorithmtoidentifycustomerpurchase getpurchase
data
fraud issuer and
sales type; and present them in to the fraud amountasshowninFigure2.
Our
techniquehasbeendesignedto forlargevolumesof complex business process data. The automatic analysis determines the importantrelationships and the visualization shows the detected relationships. contrast to Parallel Coordinatedisplays[I], werestrictthevisualizationtothree in one circulardisplayand multipledisplaysare linkedtoshowmorethan3parameters.Inaddition,weadapta newlayouttogivemoreweighttoimportantdatavalues.Figure AFraudAnalysisProcess 3.TheWeightedCircularGraphLayout
In order to give important informationmore attention, we.
&fineaweightfunctionforthenodes.Thisweightfunction allocatesmorespaceforimportantnodesandisrealizedbya weightedradian asshowninFigure3.Theweight ofanode dependsonafourthattributeA.
We&finetheweightbytheratioof attribute Aandthe s u mofallattributes
Our Approach
In this poster, we introduce a new interactivevisualization techniquetoreducedatacomplexitybyabstractingthemost critical which influencebusinessoperations.
Ourtechniqueusesthreebasiccomponents:
abusinessparametercorrelationmatrixtodetermine the
are a
(on
red, andburgundy;lessgreen)
of
thatthe of is
This might
Figure
2:Fraud Business Analysis
First publ. in: InfoVis 2004, IEEE Symposium on Information Visualization , Austin, Texas, USA , May/jun, 2004
Konstanzer Online-Publikations-System (KOPS) URL: http://www.ub.uni-konstanz.de/kops/volltexte/2008/6959/
URN: http://nbn-resolving.de/urn:nbn:de:bsz:352-opus-69591
4.2 IT Service I Analysis
4. Applications
3: processflows,-andthecauseofthe
5.
InITserviceanalysis,analysts interestedinthecauseof
unfulfilled or the impact of on
Many research efforts have focused on how to thebusiness data,asloggedby services, intovaluable Inthisapplication,the
LevelObjectives)statusindicatestheprobabilityofaservice becoming (violated),
with
0beingthemostprobable and4beingtheleastprobable. data contains10,061servicetransactionswith 50 most
criticalparametersare time,searchtime,month, day, andhour. showsITsearchtimedistribution,
Inthis wereducethecomplexityofbusinessprocess analysis by abstractingthe most critical which influencebusinessoperations.
We
theminmultiple circulargraphs.Ourreal world applications show significant of techniques in business processanalysis.A., B.: a Toolfor
Multi-Dimemiom1 CA,1990.
showa ofIT
4Ais thethreecritical andSearchtime)fromthecorrelation and are
correlatedas bynearlyparallellinesexceptsomeoutliers low to Lintswiththe
arecoloredred. 4hasthehighestsearchtime(morered,pink,burgundy)m4A.
4E
-
4H the sequenceofsearchtime (month,day,andhour)4E a circular toshowthetimedependency.The in arelinkedto bytheflowprocessmatrix. analyst
on 4in 4B 4Faregenerated,showingthat 4 with times, seenbytheblue
theanalystclickson 0in 4Cand4 0aregenerated,showingthat 0associateswithlower times,as bythe and greencolors.
4Dand4Hshowthecauseofoutliers
4D 4Haregeneratedwhentheanalystclicksonday2 searchtime-morered inFigure4E.All are out.Theanalyst onthe linesendnodes(connected day21tomonth9,and -14)todrill tothe record
as availability.
Figure4 ServiceProcess