摘要： 针对核主成分分析方法(KPCA)存在大样本集的核矩阵K计算困难问题，提出一种基于分块特征向量选择的快速核主成分分析方法。采用分块特征向量选择方法提取样本子集，用样本子集建立KPCA模型。将该方法应用于某化工过程的特征信息提取，并与全体样本的KPCA相比较。实验结果表明，两者特征提取的有效性相当，但新方法在建模和特征提取过程所耗费的时间较少。

关键词: 核主成分分析方法, 特征提取, 特征向量选择, 分块

Abstract: For a large data set, there is a problem which a kernel matrix K is not easily to be computed very mach for Kernel Principal Component Analysis(KPCA). A fast principal component analysis method is proposed to solve the computation problem for kernel matrix K based on a multi-block feature vector selection. A sample set is extracted by the multi-block feature vector selection. A KPCA model is build up by the sample. The proposed method is applied to a chemical process. Experimental result shows that the proposes method is almost same compared to the KPCA based on the all sample, but, less time are spent on modeling and extracting feature for the proposed method.

Key words: Kernel Principal Component Analysis(KPCA), feature extraction, feature vector selection, block

中图分类号: 

• TP182