基于属性加权的PCA算法
首发时间:2014-07-23
摘要: PCA是一种线性特征提取算法,通过计算将一组特征按重要性从大到小重新排列得到一组互不相关的新特征。但该算法在构造特征子集的过程中采用等权重方式,忽略了不同属性对分类的贡献是不同的。本文提出了一种把属性加权和PCA相结合的算法,通过最小化加权子空间与分类标记的距离得到各属性的权重值。得到的权重值反映了各属性对分类的贡献大小,这样生成的特征子集更有利于分类。实验结果表明,改进后的算法分类性能优于PCA算法。
关键词: 数据降维 属性加权 主成分分析(PCA) 特征提取
For information in English, please click here
An Attribute Weighted PCA
Abstract:PCA is a linear feature extraction algorithm, and calculates a set of an unrelated set of new features by dimensionality reduction. But PCA ignores that each feature of the original data attributes on classifications is different when constructing the feature subset. This paper presents an algorithm combining PCA with attribute weight, which gets the weighted values by minimizing the distance of weighted subspace and classified labels. The new feature subset is more conducive to classification. Experiment shows that the improved algorithm superiors to the PCA algorithm on the performance of classification.
Keywords: dimensionality reduction attribute weight principal component analysis (PCA) feature extraction
基金:
论文图表:
引用
No.4604439739149140****
同行评议
勘误表
基于属性加权的PCA算法
评论
全部评论0/1000