改进的模糊Min-Max神经网络与模糊系统建模

Fuzzy System Modeling Based on Modified Min-Max Neural Network

应用改进的广义模糊Min-Max神经网络进行样本分类,并以此分类结果确定模糊系统所需的模糊规则数,再运用TSK模糊系统实现函数建模,该方法的优势在于,改进的广义模糊Min-Max神经网络具有较好的自适应分类能力,可用来初步确定模糊规则数和规则空间的划分,有效避免了模糊建模时常见的规则数选取之随意性。实验结果证明,该方法实用有效。

In the paper, a new approach for fuzzy modeling is presented. The improved general fuzzy Min-Max neural network is first applied to classify the sample data set and then obtain the clusters that may be used to determine the number of fuzzy ruler and accordingly the function modeling is implemented using the famous TSK fuzzy system. The best advantage of this new approach is that the very good adaptive capability of the improved general Min-Max neural network can be utilized to initially determine the number of fuzzy ruler and the partition of the fuzzy ruler space, thus, the randomness of choosing the number of fuzzy ruler may be effectively avoided in fuzzy modeling. The experimental results demonstrate the effectiveness of this new approach.