摘要
核主成分分析(KPCA)法具有很好的非线性特征提取能力,利用KPCA提取输入数据的特征信息,并将特征信息作为最小二乘支持向量机(LSSVM)的输入变量,建立KPCA-SVM预测模型.通过实例检验表明,具有非线性特征提取的LSSVM模型的预测效果优于没有特征提取的LSSVM模型.与主成分分析(PCA)提取特征相比,KPCA特征提取效果更好.
Kernel Principle Component Analysis （KPCA） shows a very good ability for the nonlinear feature extraction. In this paper, KPCA is used to extract the nonlinear feature information, and then the obtained series are used as input of Least Square Support Vector Machine model, to establish forecasting model based on KPCA_SVM. The results show that the KPCA_SVM model is much better than that of LSSVM model without feature extraction. In comparison with PCA, the performance of KPCA was better as well.
出处
《北京师范大学学报（自然科学版）》
CAS
CSCD
北大核心
2010年第3期307-310,共4页
Journal of Beijing Normal University(Natural Science)
基金
国家自然科学基金资助项目(50779052)
