简介:ThepaperpresentsandprovesanapproximateoptimalrelationshipbetweensamplesizenandacceptancenumbercinthesamplingplansunderimperfectinspectionwhichminimizetheBayesianrisk.TheconclusiongeneralizestheresultobtainedbyA.Haldontheassumptionthattheinspectionisperfect.
简介:Artificialneuralnetworks(ANNs)havebeenwidelyusedasapromisingalternativeapproachforforecasttaskbecauseoftheirseveraldistinguishingfeatures.Inthispaper,weinvestigatetheeffectofdifferentsamplingintervalsonpredictiveperformanceofANNsinforecastingexchangeratetimeseries.Itisshownthatselectionofanappropriatesamplingintervalwouldpermittheneuralnetworktomodeladequatelythefinancialtimeseries.Tooshortortoolongasamplingintervaldoesnotprovidegoodforecastingaccuracy.Inaddition,wediscusstheeffectofforecastinghorizonsandinputnodesonthepredictionperformanceofneuralnetworks.