The Future of Analytics The Revolution Will Be Unstructured – Business Intelligence Info

The Future of Analytics The Revolution Will Be Unstructured – Business Intelligence Info

Althoughtherehasbeenmuchspeculationoverthisstatistic’sorigin,mostindustryexpertsagreethat80%to90%oftheworld’sdataisunstructureddata,andabout90%ofithasbeenproducedoverthelasttwoyearsalone。

Oftheseunthinkablyvaststores,only0。

5%iseffectivelyanalyzedandusedtoday。

Inthebusinessworld,mostunstructureddataliesincustomer-relatedtext,whichisabundantandavailable。

However,mostorganizationsdon’tknowhowtoefficientlyextractpredictiveelementsfromunstructuredcustomerdata。

They’renotsurehowtoreapthevalueoftheseinsightsbyusingthemtoboosttheperformanceofpredictiveanalytics,andmakebetteroperationalcustomerdecisions。

But,doneright,extractingvaluablepredictiveinsightsfromhugequantitiesoftexttakesjustseconds。

TheFutureIsUnstructured

Thetechindustryisfullofpredictions,butinthisone,Ihavehighconfidence:Thefutureisunstructured––becauseunstructureddataholdsthekeytothenextgenerationofintelligentsystems,whichwillbelargelybasedoncognitiveanalyticsandartificialintelligence(AI)-basedapplications。

HowdoIknow?

I’vespentmostofmycareerasadatascientistatFICO,ananalyticscompanyperhapsbestknownfortheFICO?

Score,andsoftwaresolutionsthathelpfinancialinstitutionstooptimizecreditoriginationsandfightfinancialfraud。

Allofthesesolutions,whicharepervasiveinthebankingworld,havestructureddataastheirfoundation。

Butforthepastseveralyears,I’vebeenexploringnewfrontiersforFICO,developingnovelintelligenceanalyticsacrosscomplexnetworks—theforefrontoftheanalyticsrevolution。

Specifically,myactiveresearchincludesknowledgediscoveryandbehaviorpredictionusingsupervisedandunsupervisedmachinelearningmethods,aswellasrelationallearningandnetworkprobabilisticinferencemethods。

Forthoseofyouwhoaren’tdatascientiststhattranslatesintousingunstructureddataanalysisandrelationallearningtodrivepredictiveanalyticssolutionsforapplicationsasdiverseas:

Creditoriginations

Fraud,wasteandabusedetection

Insurancepredictionandmodeling

Riskstratificationandminimization

Anticipatingofresourcedemands

Improvingmarketingcampaignsandcustomerretentionmanagementstrategies。

I’vealsobeenresponsibleforseveralexcitingproof-of-conceptinitiativesthathavebothcommercialandpublicsectorapplications。

Inalloftheseendeavors,I’vebeenexcitedbyacommonthread:thepredictivepoweroftextanalyticsandgraph-basedmethodstoimproveclassicalmachinelearningrecipes。

WhatIsTextAnalytics?

Textanalyticshelpstodeliverfreshintelligencebyminingamajorcategoryofunstructureddata,themassivestoresofcustomerdatamanyorganizationscurrentlyhaveonhand。

Hereareseveralimportantbenefits:

Discoverythroughmachinelearning:Textinemailmessages,callcenterlogs,CRMapplicationsandcollectionagentnotesisreadilyunderstoodbypeople,butit’smeaninglesstotraditionalpredictivemodels。

Machinelearningmethods,however,candeterminewhattextualdataisaboutandclassifyitforfurtheranalysis。

Thesemethodscandiscovercustomercharacteristicsandtransformthemintostructurednumericalinputsthat,inturn,canbeusedinpredictivemodelsandtraditionalanalyticalgorithms。

Handlingcomplexity:Unstructuredandsemi-structuredtext(e。

g。

XMLfiles,Excelspreadsheets,weblogs)isinherentlycomplexsinceitmaycontainawiderangeofcontentonabroadarrayoftopics。

Moreover,thepotentialvalueoftextanalysisisoftenincreasedbycombiningtextofdifferenttypesfrommultiplesources。

Suchacomprehensiveapproachmayrevealcomplex,subtlecustomerbehaviorpatternsnotevidentinsmaller,morehomogeneousdocumentsets。

Butthetaskofcollating,regularizingandorganizingdiversedatawouldbeimpossiblewithouttoday’sadvancedtechnologies。

Wenowhavetheanalytictechniquesanddatainfrastructurestoessentiallymergedisparatevarietiesoftextintoone“document”foranalysis。

Facilitatingengineering,deployment,managementandregulatorycompliance:Whiletextandtheprocessofanalyzingitcanbequitecomplex,theresultsneedtobesimpletounderstandanduse。

Todaywecanbringnewinsightsfromtextanalysisintopredictivescorecards,forexample,maintainingalltheadvantagestheyprovide。

Theresultingscorecardhashigherpredictivepower,butcanstillbeengineeredtomeetspecificbusinessneedsandregulatoryrequirements。

It’seasilydeployedintorules-drivendecisionprocessesandoperationalworkflows。

It’seasytomanage,includingtrackingperformance,automatingupdatesandmeasuringtheimpactofchange。

Thecontributionoftextanalysisistransparent,explainabletoregulators,anddocumentablethroughautomatedaudittrailsandregularvalidationandcompliancereporting。

InmynextblogI’lltakeadeeperdiveintohowtextanalyticsareusedtoboostthepredictivepowerofscorecardmodels。

Inthemeantime,ifyou’dliketolearnmoreaboutFICO’slatestresearchinunstructureddata,readmyscientificpaperpublishedbyIEEE,“MiningandVisualizingAssociationsofConceptsonaLarge-scaleUnstructuredData。

”FollowmeonTwitter@odriollet…and?

vivalarevoluciónanalítica!

Letsblockads!

(Why?

)

FICO


bck体育客服电话_bck体育_bck体育app下载