A Hybrid Model for Financial Services Social Media Sentiment Analysis Using the Bert CNN Technique

FULL TEXT in - | page 55 - 64

Abstract

The ifinancial iservices iindustry imay ifind iseveral iuses ifor isentiment iresearch, ibut iit iis inot iwithout iits ichallenges. iDue iof ithe isubjectivity iand icontext-dependence iof itext idata iinterpretation, imethods ifor isentiment ianalysis ihave iaccuracy iissues. iWith ithe iemergence iof inatural ilanguage iprocessing, isentiment ianalysis ihas ibecome ia ipowerful iinstrument ifor imeasuring ihow idiverse iindividuals ifeel iabout ia icertain iproduct ior iservice. iUtilization iof ithis itechnology ihas ibeen iadvantageous ito ithe ifinancial iservices isector. iWe iprovided ia ihybrid imodel ifor ithe ianalysis iand icategorization iof icustomers' icomments ion ifinancial iservices iduring ithe iCovid iperiod, isince iit iis iessential ithat ithey igather iconsumer ifeedback iabout itheir iservices. iWe icarried iout ian iexperiment iusing ithe iBert-CNN imodel. iThe itokenization iand iextraction iof ikey icharacteristics iwere iperformed iusing ithe iBert imodel. iThe iretrieved icharacteristics iwere iutilised ias iinput ifor ithe iCNN imodel. iThe iCNN imodel iwas itrained iusing itwelve itraining isteps, iand ithe ioutcome iindicates ithat iour iBert-CNN imodel iperforms ibetter, iwith ia itraining iaccuracy iof i98.53 ipercent iand ia itesting iaccuracy iof i96.23 ipercent. iThis ienables iour iBert-CNN imodel ito iaccurately ianalyse iand icategorise iclient ifeedback ion ifinancial iservices.

Keywords: ,,,,

References