FalconTST2.0hasachievedstate-of-the-artperformanceonMASEmetricontopglobalbenchmark.
    BanksuseFalconTST2.0tobetterforecastliquidityexposurewithBarclays,Citi,DeutscheBankandStandardCharteredhavingintegratedFalconTST2.0forcashflowforecastingandFXmanagement.
    FalconTST2.0hasachieved>93%forecastaccuracyrateconsistentlyandisreadyforbroaderindustryapplication,includinglogistics,aviation,e-commerce,besidesfinancialsector.
    SHANGHAI&SINGAPORE--(BUSINESSWIRE)--AntInternationalhasintroducedFalconTime-SeriesTransformer(TST)AIModel2.0,itsmostadvancedTSTmodelsofardesignedtodelivermoreaccurateforecastinginreal-worldFXriskmanagementofcross-borderpayments,withmoreindustryapplicationstocome,suchasdemandforecastingforsupplychainmanagementfore-commerceplatforms,andpredictiveoperationsmanagementforaviationindustry.ImplementingEnterpriseAIAndMachineLearningTools
    FalconTST2.0demonstratesState-of-the-Art(SOTA)performanceontheMeanAbsoluteScaledError(MASE)metriconatopglobalpublicevaluationbenchmarkfortime-seriesfoundationalmodels.MASEisamongthemostcriticalmetricsusedtoevaluatetime-seriesmodels.FalconTST2.0achievedaMASEscoreof0.666andplacesitatthetopoftheleaderboard,surpassingotherTSTfoundationalmodelsfromleadingglobaltechcompanies.
    FalconTSTmodelisAIbuiltforfinance
    Whilelargelanguagemodelsexcelatlearningrelationshipsintext,TSTmodelsareespeciallycriticalinfinanceandpayments,whereliquidityneeds,foreign-exchangemovements,andtransactionflowscanshiftrapidly.Thefinancialinformationconsistsofcontinuouslychangingnumericaldata—transactionamounts,accountbalances,settlementflows,andcurrencypositions.Foraglobalpaymentinstitution,theseforecastsdirectlyimpactcapitalefficiency.ThevalueofAIpredictionliesnotjustin‘calculatingmoreaccurately’butinhelpingbusinessesknowpreciselywhentheyneedfunds,howmuchtheyneed,andinwhichcurrencies.
    Thisforecastingcapabilityisequallycriticalforforeignexchangemanagement.Anairlinemaycollectticketrevenuesinmultiplecurrencieswhileneedingtopayforaircraftleases,airportfees,andoperatingcostsindifferentcurrencies.Companiestypicallyuseforeignexchangehedgingtoreducecurrencyfluctuationrisk,butthatrequiresthemtodeterminehowmuchofeachcurrencytheywillreceiveandneed.Ifforecastsaretoohigh,theymayover-hedge;iftoolow,theyleavethemselvesexposedtoforeignexchangerisk.
    Traditionalforecastingsystemstypicallybuildseparatemodelsfordifferenttasks—aretailcompanytrainsasalesforecastmodel,anairlineademandforecastmodel,afinancialinstitutionaliquiditymodel.TSTfoundationalmodelstakeadifferentapproach:FalconTSTlearnscommonpatterns—cycles,trends,seasonality,andsuddenshifts—fromdataacrossfinance,retail,energy,travel,andeconomics.Thoughtheseindustriesdiffer,theunderlyingtemporalstructuresoftensharecommonalities.
    World’sleadingbanksusingFalconTSTtoepitomiseliquidityandFXmanagement
    TheFalconTSTisinitiallydeployedinternallyatAntInternationaltomanagecashflowandFXexposureonanhourly,dailyandweeklybasis,beforehavingbeenintegratedbyleadingglobalbankstotheirownFXhedgingmodels,includingBarclays,Citi,DeutscheBankandStandardChartered,toimprovethecashflowforecastingandFXliquiditymanagementcapabilitiesforAntInternationalanditsclients.
    BarclaysintegratestheFalconTSTModelintoitsFXhedgingplatform,BARXNetFX,whileCiticombinesFalconTSTwiththeirownFixedFXRatessolution.TheyaremainlyusedforFXriskmanagementone-commerceplatformsorairlines.AndStandardCharteredusesthemodelalongsideitsSCALEFXsystemaspartofbothsides'participationinthePathFin.aiprogrammeoftheMonetaryAuthorityofSingapore.
    Currently,theyallhaveadoptedthe2.0versionoftheFalconTSTModel,leadingtoanimprovedforecastingaccuracyofmorethan93%consistently.Thislevelofprecisioniscriticalforfinancialinstitutionsmanagingvastvolumesofcross-borderpaymentsandneedingtomitigatecurrencyfluctuationriskseffectively.
    Fromasinglefinancialtooltoareusableforecastingcapability
    AntInternationalsaidanothergoalofFalconTSTistomakethesameforecastingcapabilityreusableacrosscustomersandindustries.Aviationisatypicalusecase:revenuesandcostsspanmultiplecurrencies,whilecashflowscanchangerapidly,makingaccurateforecastingcritical.FalconTSTisalreadybeingusedforFXandliquiditymanagementinthesectorandisexpandingintoe-commerce,logisticsandotherindustries.SelectingAnOnlinePaymentGateway
    Ifacapabilityservesonlyonecustomer,itremainslargelyacustomisedsolution.Whenitcanbereusedacrosscustomersandindustries,itstartstobecomeafoundationalcapability.
    “LargelanguagemodelshaveshownhowAIcanunderstandandgenerateinformation.FalconTSTisaboutanothercapabilitythatbusinessesincreasinglyneed:understandinghowtheworldchangesovertime,andanticipatingwhatcomesnext.Forus,thevalueofAIisnotsimplyachievingabetterforecastingscore,butturningthatpredictiveintelligenceintorealdecisions—howmuchliquiditytoprepare,howtomanageFXexposure,andhowtoallocatecapitalmoreefficiently.FalconTST2.0isanimportantsteptowardmakingpredictiveAIafoundationalcapabilityforglobalbusinesses,acrosspayments,accountsandbroaderfinancialservices,”saidJiang-MingYang,ChiefInnovationOfficer,AntInternational.
    Discovermore
    ApplyingForCashbackCreditCards
    SelectingAnOnlinePaymentGateway
    HiringExecutiveCoachingServices
    "FalconTSThelpsglobalbusinesses—includingourown—managecomplexcashflowandFXexposure,sotheycanmanagecross-bordertransactionswithgreaterconfidence.WithFalconTST1.0,clientssawrealoperationalvalueandcostsavingsfrombetterforecasting.WithFalconTST2.0,enhancedaccuracyandprecisionletusextendthosebenefitstoourbankingpartnersaswellasabroaderrangeofcustomersacrossfast-movingsectorslikee-commerce,travelandfintech,"saidKelvinLi,GeneralManagerofPlatformTechandSeniorVicePresident,AntInternational.
    FalconTST2.0introducesseveraltechnicalinnovationsthataddressreal-worlddatachallengestoimproveforecastingaccuracy:
    Advancedhandlingmissingdata.Nobanktransactionsoveraweekenddoesnotmeandemandiszero.FalconTST2.0distinguishesmissingdatafromactualzerovalues,reducingmisleadingpatterns.
    Powerfulgeneralisationacrossdomains.ThroughORBIT,FalconTSTlearnscommontime-seriespatternsacrossfinance,retail,energyandtourism,enablingdirectforecastinginnewbusinessscenarios.
    Supportformultipletimefrequencies.Fromsecond-levelpaymentdatatohourlytreasuryneeds,dailyairlinedemandandmonthlyeconomicindicators,FalconTST2.0handlesdifferentfrequencieswithinasinglearchitecture.
    AntInternationalinvitesinteresteddevelopersworldwidetocontributefeedbackandaccelerateinnovationintimeserieslearningthroughanAPItrialofFalconTST2.0viaGitHub.TolearnmoreabouttheFalconTSTModelandrecentsuccessstories:https://falcon-tst.ant-intl.com/
    AboutAntInternational
    AntInternationalisaleadingglobaldigitalpayment,digitisationandfinancialtechnologyprovider.Throughcollaborationacrosstheprivateandpublicsectors,ourunifiedtechfinplatformsupportsfinancialinstitutionsandmerchantsofallsizestoachieveinclusivegrowththroughacomprehensiverangeofcutting-edgedigitalpaymentandfinancialservicessolutions.Tolearnmore,pleasevisithttps://www.ant-intl.com/

　　本文链接：http://www.iruis.com/News/cninfo/76997.shtml