摘要
提出了一种基于最小分类错误率和Parzen窗的降维方法,利用Parzen窗估计数据的概率密度分布;通过计算各特征维度下的分类错误率,判断该特征维度对目标分类的贡献度;依据贡献度大小进行特征维度选择从而达到降维的目的。
A dimensionality reduction method based on minimum classification error and Parzen window is proposed, which firstly uses Parzen window to estimate the probability density of data, then calculates the contribution for classification of each feature dimension with the classification error, and selects the feature dimension according to the contribution for classification, in such a way as to achieve the intention of dimensionality reduction.
出处
《计算机工程与应用》
CSCD
2014年第14期185-188,共4页
Computer Engineering and Applications
关键词
PARZEN窗
降维
概率密度
特征选择
Parzen window
dimensionality reduction
density probability
feature selection
