神经模糊系统的估计与逼近(IJISA-V8-N6-2)

I.J.IntelligentSystemsandApplications,2016,6,9-18PublishedOnlineJune2016inMECS()DOI:10.5815/ijisa.2016.06.02Copyright©2016MECSI.J.IntelligentSystemsandApplications,2016,6,9-18EstimationandApproximationUsingNeuro-FuzzySystemsNidhiAroraITMUniverse,Vadodara,Gujarat,IndiaE-mail:[e-mail]puterApplication,Bharuch,Gujarat,IndiaE-mail:[e-mail]—EstimationandApproximationplaysanimportantroleinplanningforfuture.Peopleespeciallythebusinessleaders,whounderstandthesignificanceofestimation,practiceitveryoften.Theactofestimationorapproximationinvolvesanalyzinghistoricaldatapertainingtodomain,currenttrendsandexpectationsofpeopleconnectedtoit.Exercisingestimationisnotonlycomplicatedduetotechnologicalchangeintheworldaround,butalsoduetocomplexityoftheproblems.Traditionalnumericalbasedtechniquesforsolutionofill-definednon-linearrealworldproblemsarenotsufficient.Hence,thereisaneedofsomerobustmethodologieswhichcandealwithdynamicenvironment,imprecisefactsanduncertaintyintheavailabledatatoachievepracticalapplicabilityatlowcost.Softcomputingseekstosolveclassofproblemsnotsuitedfortraditionalalgorithmicapproaches.Toaddressthecommonproblemsinbusinessofinexactness,somemodelsareputforwardforservicing,supportandmonitoringbyapproximatingandestimatingimportantoutcomes.Thisworkillustratessomeverygeneralyetwidespreadproblemswhichareofinteresttocommonpeople.Thesuggestedapproachescanovercomethefuzzinessintraditionalmethodsbypredictingsomefutureeventsandgettingbettercontrolonbusiness.Thisincludesstudyofvariousneuro-fuzzyarchitecturesandtheirpossibleapplicationsinvariousareas,wheredecision-makingusingclassicalmethodsfail.IndexTerms—SoftComputing,Neuro-FuzzySystem,EstimationandApproximation,Decision-making,Uncertainty,Non-linearity.I.INTRODUCTIONThebusinessesintoday’seraarebecomingincreasinglycomplexandtheyarecontinuallychallengedwithdynamismofenvironment.Lotoforganizationsareexperiencingproblemsintakingunstructureddecisionsandhandlingunseeneventsoccurringeveryday,leadingtobusinessfailures.Theseunseeneventsarenotonlydynamicinnature,butalsobringalongambiguityanduncertaintywiththem.Allthereal-worldproblemsmaynotberepresentedproperlyusingtraditionalapproachesduetothelackofpreciseknowledge,theirnon-linearbehaviorortheirhighdegreeofuncertainty.Inanattempttosimplifythecomplicationsofthereal-world,thesemodelstendtooverlooktheactualbehaviorofbusiness,whichmaketheminefficienttouseandsometimesdoesnotgivethedesiredresults.Manytechniquesrangingfromregressionanalysistotimeseriesareavailableandbeingregularlyusedbyorganizationstogenerateforecasts,butthemethodscurrentlyinusesomewhereareincapableofpredictionoffutureevents.Asystematicconsiderationoffuzzydatainmanagingcomplexissuesoftodayandpredictingtheuncertainissuesoftomorrowcanonlyhelpthemsucceed.Theaimofthisworkistoproposesolutionstosomereal-worldproblemswhichsufferfromuncertaintyandvaguenesswiththeintentiontostudythemethodscurrentlyinpractice.Throughproperstudyandobservation,theidentificationofuncertainattributesandthefuzzinessinvolvedineachproblemisdoneandsubsequentlyhandledwithapplicationofNeuro-fuzzymethodologies,wherehugedatacanbehandledbytrainabilityandadaptationofneuralnetworkandfuzzylogicisthewayouttovagueness.Anattemptismadetofindouthowthehybridsoftcomputingapproaches,particularlycombinationofdifferenttypesofneuralnetworkwithfuzzylogic,canbeappliedforthepurposeofestimationandapproximationinreal-worldproblems.Accuracyhasbecomedreamboatforresearchers,butinthehuntforaccuracytheysometimesignoreimportantthings.AccordingtoFortuna[Fortuna,2001][1]thebasicprincipleofsoftcomputingisitscombineduseofnewcomputationtechniquesthatallowittoachieveahighertoleranceleveltowardsimprecisionandapproximation.Asopposedtoconventionalmethods,softcomputingmethodologiesmimicconsciousnessandcognitioninseveralwayslikelearningfromexperience,performinginput-outputmappingetc.bysimulatingbiologicalprocessthroughparallelization.Estimationormorespecifically,predictionisanartwherebusinessleaderstrytominimizeuncertaintybyidentifyingandevaluatingtheassociatedrisk.Thetermspredictionandforecastingareusedinbusinessexchangeablyandsoishere.Abusinesswhichcanadjustitselfwiththechangingneedsofmarketanditscustomerscanhaveabright10EstimationandApproximationUsingNeuro-FuzzySystemsCopyright©2016MECSI.J.IntelligentSystemsandApplications,2016,6,9-18futureandcangolong.Statisticaltechniquesaremostlyusedbybusinesspeopleforcalculatingatrendorforhavinganinsightinfuture.Accuratepredictionsallowbusinessleaderstobereadyforfutureupsanddownsbyforetellingthefuturewithprecision.Tosurviveinthecompetitivemarketandtobeabletofulfillexpectationshavebecomecrucialforthebusinesspeople.Apparently,thiscannotbeachievedwithouteffectivelyforeseeingsomeimportantfutureevents.Thisinturngivesthemchancetoattractnewcustomers,retainexistingcustomers,understandpresentscenarioandpossiblysolveproblemswhichmayoccurinfuture.Now-a-dayssoftcomputingmethodologiesaredeployedasanalternativeto