Abstract: In allusion to the limitation of Principal Component Analysis （PCA） in solving the multiple correlations between variables，this paper presents an improved PCA feature extraction method based on K-maxmin clustering，which combines with RelieF algorithm to remove non-correlative features for improving efficiency and accuracy farther.Experimental results show that this method is better than traditional PCA method in feature extraction.

Key words: feature extraction, Principal Component Analysis（PCA）, multiple correlation, RelieF algorithm, K-maxmin clustering

摘要： 针对主成分分析（Principal Component Analysis，PCA）在克服变量多重相关性中的局限作用，提出了基于K-maxmin聚类的改进PCA特征提取方法，并结合RelieF算法去除分类不相关特征，可进一步提高算法效率和准确性。实验结果表明，该方法的特征提取效果优于传统的PCA方法。

关键词: 特征提取, 主成分分析, 多重相关, RelieF算法, K-maxmin聚类