采用主成分分析法 (PCA)来改善径向基函数网络的泛化性能 ,理论上可以根据PCA方法中的主成分累积贡献率 ηK 决定RBF网络的输入层节点数 .实例研究证明 ,采用PCA方法后的RBF网络泛化性能良好 PCA is used to improve the generalization performance of radial basis function networks. Theoretically, the number of input layer nodes in RBF network can be determined theoretically according to the cumulative contribution rate ηK of principal components in PCA method. Case studies show that PCA method After RBF network generalization performance is good