模糊逻辑系统的区域分割学习方法

Fuzzy Rule Region-split Method for Fuzzy Logic System

模糊逻辑系统的区域分割学习方法采用规则前件为不对称高斯型隶属度函数的模糊逻辑系统 .新规则生成的判据是模糊逻辑系统对新样本数据的逼近性能达不到期望逼近精度 .方法定义了规则有效范围的概念 ,并假定模糊逻辑系统中各规则的有效范围不能互相冲突 ,在输入空间内的一个点只能位于唯一一条规则的有效范围之内 ,并据此来确定新规则的有效范围 ,新规则的有效范围是由原所在规则的有效范围分割而来 .方法的应用前提条件很少 ,只需要事先确定期望逼近精度和有效范围系数 .

Fuzzy rule region-split method adopts the fuzzy system whose antecedents take the asymmetric Gauss function as membership function.The guideline to generate a new fuzzy rule is that the approximation performance of FLS can't reach the desired precision for the new data pair.The method defines the concept of the effective range of fuzzy rule and assumes that each rule's effective range should not overlap other rules' and an arbitrary point in the input space can only be located in the effective range of a fuzzy rule.The new rule's range is derived from splitting the range of the original rule.The method has two preconditions,the desired precision and the effective range constant,and thus it's very oriented to the industrial application.