资源描述:
《公司财务困境预测 外文翻译资料》由会员上传分享,免费在线阅读,更多相关内容在行业资料-天天文库。
1、AnApplicationofSupportVectorMachinetoCompani’sFinancialDistressPredictionBecauseoftheimportanceofcompaniesfinancialdistressprediction,thispaperappliessupportvectormachine(SVM)totheearly-warningdistress.Takinglistedcompaniesthree-yeardatabeforespecialtreatme
2、nt(ST)assampledata,adoptingcross-validationandgrid-searchtechniquetofindSVMmodel’sgoodparameters,anempiricalstudyiscarriedout.BycomparingtheexperimentresultofSVMwithFisheranalysis,Logisticregressionandbackpropagationneuralnetworks,itisconcludedthatfinancial
3、distressearly-warningmodelbasedonSVMobtainsabetterbalanceamongfittingability,generalizationabilityandmodelstabilitythantheothermodels.Stockholdersandotherinterestpartssufferindividualeconomicloss,butalsoifmanyenterprisesrunintobankruptcytheeconomicdevelopme
4、ntofthewholecountrywillbegreatlyshocked.Ingeneral,mostenterprisesthatranintobankruptcyhadexperienceconditionoffinancialdistress,buttheycouldnotdetectfinancialdistressatanearlystageandtimelytakeeffectivemeasurestopreventbankruptcy,intheperspectiveofmanagemen
5、t,itisimportanttoexploreamoreeffectivefinancialdistresspredictionmodeltosignalearly-warningforenterpriseswhichwillpossiblygetintofinancialdistress,sothatmanagerscantakestrategicactionstoavoiddeteriorationoffinancialstateandbankruptcy,fromtheviewoffinanciali
6、nstitution,aneffectivefinancialdistresspredictionmodelcanhelpthemdetectcustomerswithhighdefaultriskatanearlystagesoastoimprovetheirefficiencyofcommercialcreditassignment.Beaver(1966),oneofthefirstresearcherstostudybankruptcyprediction,thepredictabilityofthe
7、14financialratiosusing158samplesconsistingofailedandnon-failedfirms[2].Beaver’sstudywasfollowedbyAltman’smodel(1968)basedontheMDAtoidentifythecompaniesintoknowncategories.AccordingtoAltman,bankruptcycouldbeexplainedquitecompletelybyusingacomboffive(selected
8、fromanoriginallistof22)financialratios.Logmodelisusedtodealwithtwoclassesclassificationproblems,andOhlsonwasthefirstapplyittopredictingfinancialdistressin1980[4].Themostwidelyusedmachinelearningmethodi