利用神经模糊系统从数据中挖掘模糊逻辑

Fuzzy Rules from Data Mining Based on Fuzzy - Neural Systems

提出一种利用神经模糊系统从实验记录的曲线或者数据中挖掘模糊逻辑规则的方法。首先根据生产控制系统的实验记录的曲线或者数据,初步确定一个“if…then”模糊逻辑规则模型;然后利用具有非线性特性的神经网络和模糊推理中的模糊逻辑运算,构造了一个神经模糊系统;通过有导师的误差反向传输学习,将训练后的神经模糊系统的网络联接权的变化结合为模糊逻辑规则的变化和修改,从而实现了从实验记录的曲线或者数据中推理、归纳的模糊逻辑控制规则。

An approach of fuzzy rules from data mining on fuzzy - neural systems (FNS) is presented in the paper. A rude model of if-then fuzzy logic inference is defined based on real experimental consequent data firstly. A fuzzy - neural system is proposed by means of a neural network incorporating fuzzy logic inference. Updating weights of the fuzzy - neural system is taken into account of revision of fuzzy logic rules by a BP learning procedure. The approach proposed not only accommodates fuzzy logic inference rules from real experimental consequent data, but realizes more efficiently to control real life system in fully parallel by FNS.