摘要
为解决粗糙集离散化过程中存在的信息损失问题,将粗糙集理论与模糊集理论相结合,提出基于FCM的模糊粗糙属性约简算法。该方法用模糊C均值聚类算法对连续属性进行模糊化,并通过有效性分析来确定最佳分类数目。该方法克服了目前属性模糊化方法需要人为规定划分类数,几乎不考虑信息系统的具体属性值等缺点。最后分别对天气信息系统和玻璃识别信息系统进行了属性约简计算,结果表明该方法是可行有效的。
In order to solve the problems of information loss in the course of attribute discreted based on rough set,a new method of attribute reduction based on fuzzy rough set and fuzzy clustering method is proposed. In view of the deficiency of attribute fuzzified method,the fuzzy C means clustering is introduced in order to fuzzify the continual attribute, and the best minute class number is obtained by the valid analysis of clustering. Attribute reductions of weather and glass identification information systems are improved to prove feasibility and effectiveness of this method.
出处
《现代电子技术》
2009年第17期194-196,共3页
Modern Electronics Technique
关键词
模糊聚类
模糊粗糙集
决策表
属性离散化
fuzzy clustering
fuzzy rough set
decision table
attribute discretization
