摘要: 针对由类的重叠引起的训练样本模糊不确定性,以及属性不足引起的类边界粗糙不确定性,提出一种基于期望-最大化(EM)的模糊-粗糙集最近邻分类算法——EM-FRNN。利用UCI数据库的突发性水污染事件案例进行实验,实验结果表明,与朴素的KNN、模糊最近邻算法、模糊粗糙最近邻算法相比,该算法的运算精度高且计算成本较低。
关键词:
最近邻,
模糊-粗糙集,
期望-最大化,
EM-FRNN算法
Abstract: For fuzzy-uncertainty with class overlap and rough-uncertainty with lack of features, this paper proposes a fuzzy-rough nearest neighbor clustering classification algorithm based on Expectation-Maximization(EM), named EM-FRNN. Through the experments with UCI emergency water pollution cases database, compared with the classic algorithms, such as KNN, FKNN, FRNN, EM-FRNN algorithm improves classification precise and reduces computation.
Key words:
nearest neighbor,
fuzzy-rough set,
Expectation-Maximization(EM),
EM-FRNN algorithm
中图分类号:
何力, 卢冰原. 基于EM的模糊-粗糙集最近邻算法[J]. 计算机工程, 2010, 36(24): 136-138.
HE Li, LEI Bing-Yuan. Fuzzy-rough Set Nearest Neighbor Algorithm Based on EM[J]. Computer Engineering, 2010, 36(24): 136-138.