Abstract: How to decrease the time of training and testing the samples，and improve the classification accuracy are important aspects of the feature selection research.A new feature selection approach by PCA and ReliefF is presented in this paper.The algorithm can take out the most representative features which constitute the effective feature sets from the original features，thus the dimensions of the features are decreased.Moreover the algorithm is proven to be more advantageous than the approach of PCA-GA in its simplicity and speed.Experiments on a UCI dataset show that the method in this paper provides a new research approach for information feature compression in pattern recognition.

Key words: feature selection, principle component analysis, support vector machine

摘要： 如何减少样本的训练测试时间、提高分类精度是有效特征选择方法研究的重要方面。提出了一种结合PCA和ReliefF的特征选择算法。该算法选择出了最具有代表性的特征，构成有效特征子集，实现了特征降维。同时，较PCA-GA方法，该算法具有简单、快速等优点。利用标准数据集进行的实验结果表明，文中算法是可行的、有效的，为模式识别的信息特征压缩提供了一种新的研究方法。

关键词: 特征选择, 主成分分析, 支持向量机