利用人工神经网络观测社交网络人个性研究(IJMECS-V5-N8-6)

I.J.ModernEducationandComputerScience,2013,8,51-57PublishedOnlineOctober2013inMECS()DOI:10.5815/ijmecs.2013.08.06Copyright©2013MECSI.J.ModernEducationandComputerScience,2013,8,51-57ArtificialNeuralNetworkinPrognosticatingHumanPersonalityfromSocialNetworksHarishKumarVPESITSouthCampus,Bangalore,[e-mail]pus,Bangalore,[e-mail]hnologies,Bangalore,[e-mail][e-mail]—Theanalysisoftextintheformoftweets,chatorpostscanbeaninterestingaswellaschallengingareaofresearch.Inthispaper,suchananalysisprovidesinformationaboutthehumanbehavioraspositive,negativeorneutral.Forsimplicity,tweetsfromsocialnetworkingsite,Twitter,areextractedforanalyzinghumanpersonality.Variousconceptsfromnaturallanguageprocessing,textminingandneuralnetworksareusedtoestablishthefinaloutcomeoftheapplication.Foranalyzingtext,NeuralNetworksareimplementedwhicharesomodeledthattheypredicttheHumanbehavioraspositive,negativeorneutralbasedonextractedandpreprocesseddata.UsingNeuralNetworks,theparticularpatternisidentifiedandweightsareprovidedtowordsbasedontheextractedpattern.Neuralnetworkshaveanaddedadvantageofadaptivelearning.Thisapplicationcanbeimmenselyusefulforpolitics,medicalscience,sports,matrimonialpurposesetc.Theresultssoobtainedarequitepromising.IndexTerms—NeuralNetworks,SocialNetworkTextAnalysis,TextMining,WordnetI.IntroductionThereisavastpoolofunstructuredtextdataavailableonInternetthatisincreasingexponentiallydaybyday.Someofthemostpopularsocialnetworkingsitesarethegreatsourcesofsuchunstructuredtextdata.Thispaperpresentsausefulwayofanalyzingsuchdataformanydecisionmakingpurposes.Suppose,acompanywantstohiresomeprofessionalsandatthesametimetheywanttoknowaboutthebasicpersonalitytraitsofthepeopleunderconsideration.Insuchasituation,theapplicationproposedinthispapercanhelprevealingwhetherthepersonisoptimistic(positive),pessimistic(negative)orneutral.Suchinformationcanhelpthehiringcompanytodecidewhetherthepersonwouldbeanassetfortheorganizationornot.Similarly,whenparentslookforamatchfortheirsonordaughter,theycanusethisapplicationtogetaninsightintothebasicattitudeoftheperson.Here,anassumptionismadethatthepersontobeanalyzedmusthaveanaccountonTwitter(thesocialnetworkingsiteconsideredfortheresults).So,thiswaytheproposedapplicationcanhelptakingveryimportantdecisionsincrucialscenarios.Thetextisextractedandpreprocessed(cleaned)forfurtheranalysis.Foranalyzingtext,themultilayeredartificialneuralnetworksareimplemented.TheconceptofNeuralNetworksisinspiredfromthehumanbrainandnervoussystem.Aneuralnetworkconsistsofinformationprocessingunitcalledaneuron.Neuralnetworksarequiteeffectiveandcompetentwithhugedata.Theyareusuallyusedtomodelintricaterelationshipsbetweeninputsandoutputstofindunderlyinghiddenpatternsintextualdata.NeuralNetworkshavetheabilitytoadapttomodifiedinputsothatthenetworkproducesresultwithouttheneedtoredesigntheoutputcriteria.ANeuralNetwork[1]iscategorizedbasedonmanyaspects.Someoftheaspectsare,itspatternofconnectionbetweentheneuronscalleditsarchitecture,itsmethodsdeterminingtheweightsontheconnection52ArtificialNeuralNetworkinPrognosticatingHumanPersonalityfromSocialNetworksCopyright©2013MECSI.J.ModernEducationandComputerScience,2013,8,51-57calleditstrainingandlearning,algorithmsanditsActivationFunction.Also,thedesignofthesenetworksisparallel.Evenifoneneuronisnotworking,thenalsotheclassificationresultsareobtained.Also,TextMiningisaprocessoffullorpartialautomationofexploringunknownhiddenpatternsinthetext.Itextractsrelevantinformationfromtext,whichisotherwisenotveryobvious[2].Inthispaper,texthasbeenputtopreprocessingsteptomakeitappropriateforfurtheranalysis.Thevariousstopwordsareremoved.Thecommonwordsthatoccurquitefrequentlyintextareunlikelytohelpminethetext.E.g.,“the”,“a”,”an”aresomeofthestopwordswhicharebeingremoved.Thewordslike“he,“she”,“we”,“them”,“him”etc.arenotremovedasstopwordsbecausesuchwordsgivetheinformationaboutcertainimportantfeaturesregardingthepersonunderconsideration.Thecleanedtextisanalyzedbyusingvarioussimilaritymeasuresandsemanticsimilarity.Intextmining,NaturalLanguageProcessingalsoplaysavitalrole.NaturalLanguageProcessing(NLP)[3]isacomputerizedapproachtoanalyzetextbasedonvarioustheoriesandtechnologies.Althoughitisaveryactiveareaofresearchanddevelopment,thereisnosingledefinitionthatwouldsatisfyeveryone.TheapplicationproposedinthispaperisbestsuitedforthefollowingfieldsbutnotlimitedtoCriminalSciences,MedicalScience,InsuranceIndustrySportsHumanResourceDepartmentForMatrimonialpurposesPoliticsetc.Thetextisextractedfromsocialnetworkingsite,Twitterandpreprocessedbyremovingsomestopwords.Oncetheinputtextisready,itisfedtotheclassifierandclassifierclassifiesthepersonaspositive,negativeorneutral.Hereclassifierisdevelopedusingmultilayeredartificialneuralnetworks.Themotivationforimplementingartificialneuralnetworksforthepurposeisthatitadaptstotheever-changinginputandaccordinglyprovidestheo