摘要： 提出了一种新的模糊规则提取方法,该方法先采用基于山峰函数的减法聚类法自适应地确定初始的聚类中心,然后由此构造动态自组织神经网络进行学习,在学习的过程中可根据情况适当地合并或分裂神经元,并重构神经网络继续学习,最后按聚类中心确定模糊子集数目和隶属函数并形成模糊规则集。实验结果表明,通过网络结构和神经元的动态自适应变化能够获取样本集中的模糊信息,形成直观的模糊规则。

关键词: 模糊规则提取, 自组织特征映射, 减法聚类法

Abstract: A new method for extracting the fuzzy rules is proposed, which uses subtraction clustering based mountain function to automatically find out the original clustering centers. IT also can construct a dynamic neural network for learning. During the training , the network can automatically divide or incorporate nerve cell ,and then reconstructs itself to learn again. Finally, on the basis of clustering centers , it can acquire the membership functions and fuzzy rules. The experimental results show that the method can acquire the fuzzy information of sample set and extract the fuzzy rules.

中图分类号: 

• TP18