摘要
提出将Rough集理论与构造型神经网络覆盖算法相结合,用Rough集理论提取保持信息完整的最小属性集后构造覆盖网络,提高了覆盖算法的泛化能力,而对于属性不完备信息系统进行粒度处理后再构造覆盖网络,能解决覆盖算法对不完备信息系统的分类.实验结果表明该算法能提高覆盖算法的应用范围和对不完备信息系统的知识发现.
This article attempts to the utmost extent to solve the classification problem based on the Rough Set Theory and Structural Covering Neural Networks. The covering domain will be constructed after the minim reducing attributes are extracted from the completing samples in the guide of rough set theory, which improves the generalization of the covering algorithm. At the same time the data of incomplete information system are processed to consistent decision data in granularity study, which will solve the classification problem of the incomplete information system. The experiment results show that it can largely enlarge the application of the structural neural networks covering algorithm and discover more knowledge of the incomplete information system.
出处
《模式识别与人工智能》
EI
CSCD
北大核心
2004年第3期362-367,共6页
Pattern Recognition and Artificial Intelligence
基金
国家自然科学基金对外合怍项目(No.60111120662)
安徽省教育厅自然科学研究基金(No.2002jk002)
关键词
构造型神经网络
覆盖算法
属性约简
粗糙集
Structural Neural Network
Covering Algorithm
Attribute Reduction
Rough Set
