WEBKDD 2000 - Web Mining for E-Commerce.

Full text

SIGKDDExplorations.  Volume2,Issue2–page106
WEBKDD2000-WebMiningforE-Commerce
MyraSpiliopoulou
Otto-von-GuerickeUniversitaet.
Universitaetsplatz2,D-39016
Magdeburg
myra@iti.cs.uni-magdeburg.de
JaideepSrivastava
UniversityofMinnesota&Yodlee,Inc
3600BridgeParkway,Suite200
RedwoodCity,CA94065
srivasta@cs.umn.edu
BrijM.Masand
RedwoodInvestmentSystems
76SummerStreet
Boston,MA02110-1225USA
brij@redwood.com

RonKohavi
BlueMartiniSoftware
2600CampusDrive,SanMateo,
California94403
ronnyk@bluemartini.com


ABSTRACT
Inthispaper,weprovideasummaryoftheWEBKDD2000
workshop,whosethemewas‘WebMiningforE-Commerce’.
ThisworkshopwasheldinconjunctionwiththeACMSIGKDD
InternationalConferenceonKnowledgeDiscoveryinDatabases
(KDD-2000).
Keywords
Webmining,e-commerce,personalization,clickstreamanalysis.
1. THEME

The ease and speed with which business transactions can be
carriedoutovertheWebhasbeenakeydrivingforceintherapid
growth of electronic commerce. In addition, customer
interactions, including personalized content, e-mail campaigns,
andonlinefeedbackprovidenewchannelsofcommunicationthat
werenotpreviouslyavailableorwhereveryinefficient.TheWeb
is revolutionizing the way businesses interact with each other
(B2B)and witheachcustomer(B2C).Ithasintroducedentirely
newwaysofdoingcommerce,includinge.g.auctionsandreverse
auctions. It also made it imperative for organizations and
companiestooptimizetheirelectronicbusiness.

Knowledge about the customer is fundamental for the
establishmentofviablee-commercesolutions.Webminingfore-
commerceistheapplicationofwebminingtechniquestoacquire
thisknowledgefore-commerce.Typicalconcernsine-commerce
includeimprovedcross-sells,up-sells,personalized ads,targeted
assortments,improvedconversionrates,andmeasurementsofthe
effectivenessofactions.

TheWEBKDD 2000 workshopis thesecondworkshop held in
conjunctionwiththeACMSIGKDDInternationalConferenceon
KnowledgeDiscoveryin Databases(KDD)anddedicatedtothe
challengesofwebmining.WEBKDD'99focusedontheaspectsof
web mining related to user profiling; the long version of its
proceedings has appeared as volume 1836 of the Lectures in
Artificial Intelligence series (LNAI) by Springer Verlag. In
response to call for papers, WEBKDD 2000 received 31
contributions. Each was reviewed by at least three program
committee members. Seven submissions were selected for
presentationaslong papers,andsixasshortpapersreportingon
goodideasataratherpreliminaryphase.

The URL http://robotics.stanford.edu/~ronnyk/WEBKDD2000
containsthe finalversionsof theworkshoppapersandtheslide
presentations.

2. WORKSHOP

The KDD community responded very enthusiastically to the
WEBKDD2000workshop,andwereceivedfarmorerequestsfor
attendance(approximately110)thanthere wasspace.About85
people attended the workshop, which brought together e-
commercepractitioners,toolvendorsanddataminingresearchers.
Thepaperpresentationwasdividedintothreesessions.

The first session, titled Web personalization and recommender
systems, focused on how web mining can address one of the
fundamental issues of B2C e-commerce, namely personalized
customerexperience. Asoftendescribed byJeffBezos, CEOof
Amazon.com, and mentioned by Joseph Pine in his The
Experience Economy [1], customer experience is the key to
building customer loyalty to an on-line store, since leaving the
storeisexactlyoneclickaway.Inthissessionwehadthreelong
andtwoshortpapers, whichpresentedtheleadingedgeideasin
this important area. Mobasher, Dai, Luo, Nakagawa, Sun, and
Wiltshire’spaper,titled Discovery ofAggregateUsage Profiles
forWebPersonalization,describedhowusagedatafromweblogs
canbeanalyzed/minedtobuilduserprofiles,andhowthesecould
be use to enhance the user’s browsing experience.Vucetic and
Obradovic’s paper, titled A Regression-Based Approach for
Scaling-UpPersonalizedRecommenderSystems inE-commerce,
presented an approach to applying regression techniques to
understand user preferences for recommender systems. This
approachisinterestingsincestatisticaltechniqueshavenotbeen
appliedsufficientlytothisproblem.Sarwar,Karypis,Konstanand
Riedl’spaper, titled Applicationof DimensionalityReductionin
Recommender Systems – A Case Study, presented a novel
applicationofdimensionalityreductiontechniquesfromscientific
computingtotherecommendationsystemproblem.Thetwoshort
papers in this session, namely Lin, Alvarez and Ruiz’s
Collaborative Recommendation via Adaptive Association Rule
MiningandChangandYuan’sASynthesizedLearningApproach
for Web-Based CRM, presented early results in alternative
approachestotherecommendationsystemproblem.Insummary,
the wealth of interest in applying various techniques to the
recommendation system problem shows the centrality of this
problemtoWebpersonalization.

SIGKDDExplorations.  Volume2,Issue2–page107

Thesecond session,titledMining frameworksandcase studies,
presented experience reports from four groups on using web
mininginvariouse-commerceapplications.Anyfieldofenquiry
musthaveits‘proofofthepuddingisineatingit’component,and
thecasestudiespresentedhereprovideexactlythatflavor.Ansari,
Kohavi, Mason and Zheng’s paper, titled Integrating E-
Commerce and Data Mining: Architecture and Challenges,
providedacomprehensiveoverviewoftheissuesinapplyingdata
miningtechniquestoE-commerce.Theauthorsbringalotofreal
worldperspectivefromtheirexperienceatBlueMartiniSoftware,
especially from the viewpoint of an E-commerce solution
provider.TheusingerandHuber’sAnalyzingthefootstepsofyour
customers, A case study by ASK|net and SAS Institute GmbH
presentedtheexperienceofapplyingdatamininginE-commerce
from a solutions perspective, where SAS’s tools were used to
solve a problem for ASK|net. Sanford Gayle, in his paper The
MarriageofMarketBasketAnalysistoPredictiveModeling:The
EssentialChallenge inExploitingWeb-LogFilesforPrediction,
presented an approach to using association and correlation
analysis to extract predictive models from web logs. Coenen,
Swinnen, Vanhoof and Wets’ paper A Framework for Self
AdaptiveWebsites:TacticalversusStrategicChallengeexamined
the various issues in building such sites. The essential tension
seemstobebetweenmakingawebsitepersonalizedtoindividual
users-maybeevendynamicallychangeitbasedontheparticular
user’sbehavior–andtheinformationoverloaditcancause.

The third session, titled Navigation analysis, focused on how
clickstreamdata canbeanalyzedtoextractvaluablee-commerce
knowledge from it. Being able to analyze clickstream data
providesanunprecedentedopportunitytounderstandindetailthe
processleadinguptoabuy/notbuydecisionvs.justrecordingthe
finaloutcome-asisthecasewithpoint-of-saledata.Clickstream
data is over 95%+ of all data collected in most large-scale e-
commerce environments, and contains a wealth of knowledge
embeddedin it.Berendt’spaper, titledWebUsageMining, Site
Semantics, and the Support of Navigation, provided a general
overviewoftheissuesinclickstreamanalysis,andhowthemined
knowledge can be used for supporting site navigation. Kato,
Nakayamaand Yamane’spaper NavigationAnalysis Toolbased
on the Correlation between Content Distribution and Access
Patterns,presentedanapproachwherebyminedpatternsfromsite
contentcanbecorrelatedwithminedpatternsfromsiteusage,and
atoolbasedonthisapproach.Investigatingapplicationsofsucha
toolwouldbe aninteresting lineof inquiry.Gaul andSchmidt-
Thieme’s Mining Web Navigation Path Fragments presented
somenovelalgorithmsforextractingnavigationalpathfragments.
Finally,TanandKumar’sModeling ofWebRobotNavigational
Patterns addressed the challenging and commercially important
problemofseparatingthesitevisitsofwebrobotsfromhumans.
This is crucial for at least two applications: (1) as competitive
pressuresincrease,commercesiteswouldliketoblockrobotsthat
collectsensitiveinformation,and(2)accuratemodelingofhuman
users’e-commercebehaviorrequiresthatwebrobotaccessesbe
filteredout.Whilethepapersinthissessionpresentsomeofthe
leadingideas, theresearchinthisareaisjust beginningand we
havebarelyscratchedthesurface.
3. CONCLUSION

WEBKDD2000turnedouttobeaverysuccessfulworkshopby
all measures. More than 110 people showed interest in the
workshop and over 85 attended it. The quality of papers was
excellent,thediscussionwaslively,and anumberof interesting
directions of research were identified. This is a strong
endorsementofthelevelofinterestinthisrapidlyemergingfield
ofinquiry.
4. REFERENCES

[1] B. JosephPine,James H.Gilmore, B.JosephPineII. The
Experience Economy. Harvard Business School Pr; ISBN:
0875848192,
http://www.amazon.com/exec/obidos/ASIN/0875848192/ref
=sc_b_1/103-2009916-9046229

