摘要
如何从大量数据中提取或“挖掘”知识是数据挖掘领域中的研究热点之一。利用神经网络在提取规则时的优势从样本数据中提取模糊规则。介绍了一系列算法 ,其中朴素提取规则仅是对样本数据粗糙的挖掘。参照模糊控制中的模糊化思路 ,可在挖掘问题中引入模糊语言变量。根据隶属函数的训练 ,模糊语言的筛选 ,属性间是否存在相关性的判断等问题 ,利用神经网络中的BP算法提出了双向训练算法。在已完成训练的网络进行网络剪裁 ,最后在完成剪裁的网络上 ,先确定候选规则再利用聚类结果从候选规则中提取模糊规则。
ion algorithm could only dig sample data roughly. To make reference to fuzzy method of fuzzy control, fuzzy language variables were quoted. To base some questions about training subjection function, selecting fuzzy language and confirming pertinence among attributes and so on, utilized BP algorithm of neural network, double-way training algorithm was presented. Then the network was cut out which had trained. Finally, on the network which had been cut out, by using clustering result, fuzzy rule was extracted from the decided alternative rule.
出处
《石油化工高等学校学报》
CAS
2004年第3期83-88,共6页
Journal of Petrochemical Universities
关键词
数据挖掘
规则提取
模糊规则
神经网络
算法
Data mining
Rule abstraction
Fuzzy rule
Neural network
Algorithm