摘要
IntheresearchofusingRadialBasisFunctionNeuralNetwork(RBFNN)forecastingnonlineartimeseries,weinvestigatehowthedifferentclusteringsaffecttheprocessoflearningandforecasting.Wefindthatk-meansclusteringisverysuitable.Inordertoincreasetheprecisionweintroduceanonlinearfeedbacktermtoescapefromthelocalminimaofenergy,thenweusethemodeltoforecastthenonlineartimeserieswhichareproducedbyMackey-Glassequationandstocks.Byselectingthek-meansclusteringandthesuitablefeedbackterm,muchbetterforecastingresultsareobtained.
出版日期
2003年02月12日(中国期刊网平台首次上网日期,不代表论文的发表时间)