简介：ThepaperpresentsandprovesanapproximateoptimalrelationshipbetweensamplesizenandacceptancenumbercinthesamplingplansunderimperfectinspectionwhichminimizetheBayesianrisk.TheconclusiongeneralizestheresultobtainedbyA.Haldontheassumptionthattheinspectionisperfect.

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