Clustering： from Clusters to Knowledge.pdfVIP

ComputerTechnologyandApplication4(2013)284-290 lN塌 Clustering：frOm ClusterstoKnowledge PeterGrabusts ComputerScienceandMathematicsDepartment,FacultyofEngineering,RezekneHigherEducationalInstitution，RezekneLV-4601 Latvia Received：May17，2013／Accepted：June18，2013／Published：June25，2013 Abstract：Dataanalysisandautomaticprocessingisofteninterpretedasknowledgeacquisition．Inmanycasesitisnecessaryto somehow classify dataorfindregularitiesin them．Resultsobtained inthesearchofregularitiesin intelligentdataanalyzing applicationsaremostlyrepresentedwiththehelpofIF-THEN rules．W iththehelpoftheseurlesthefollowingtasksraesolved： prediction，classification，patternrecognitionna dothers．Usingdifferentapproachesclusteringalgorithms，neuralnetworkmethods， fuzzyurleprocessingmethOds_ wecanextractrulesthatinna understnadablelnaguagechraacterizethedata．Thisallowsinterpreting thedata，findingrelationshipsinthedataandextractingnewrulesthatcharacterizethem．Knowledgeacquisitioninthispaperisdefined astheprocessofextractingknowledgefromnumericaldatainthefo1111ofrules．Extractionofurlesinhtiscontextisbasedonclustering methodsK-meansnadfuzzyC-menas．WihthteassistanceofK-means．clusteringalgorithm rulesraederivedrfom trainedneural networks．FuzzyC．meansisusedinfuzzyurlebaseddesignmethod．Ruleextractionmethodologyisdemonstratedintl1eFisher’sIris flowerdatasetsamples．Theeffectivenessoftheextractedurlesisevaluated．Clusteringnadurleextractionmethodologycna bewidely usedinevaluatingnadanalyzingvraiouseconomicnadfinancialprocesses． Keywords：Datanaalysis，clusteringalgorithms，K-means，fuzzyC-mena s，urleextraction 1．Introduction characterizesIDA asaprocesswhich “findsinthedata regularities previously unknown， non—trivial， TheaimofIDA (intelligentdataanalysis)istofind