摘要
KNN是一种简单、有效、非参数的分类算法.针对样本分布偏斜的分类环境,首先提出了一种改进的特征选择方法进行特征降维,在此基础上进一步提出了一种基于分布的改进KNN方法用于文本分类,降低了分布偏斜问题对决策函数的影响.试验表明,所提出的改进KNN文本分类方法具有较好的分类性能.
KNN is a simple, valid and non-parameter method often applied in categorization. Under the condition that the samples distribution is uneven, we first put forward an improved weighting way in feature reduction; then we improve the KNN basing on density in automatic text categorization. This way reduces the impact from the uneven distribution, we have a test about text categorization. The result shows that these methods have a better precision than the common KNN.
出处
《微电子学与计算机》
CSCD
北大核心
2010年第3期51-53,58,共4页
Microelectronics & Computer
基金
国家自然科学基金项目(70571087)
关键词
特征选择
文本分类
改进KNN
相似度
feature selection
text classification
improved KNN
similarity