An efficient method of fuzzy rules generation

This paper proposes a method for the automatic generation of fuzzy rules that finds out the essential points of the control surface by the concept of K-nearest-neighbor, and then uses these points to determine the fuzzy partitions so that it can construct a fuzzy neural network to learn fuzzy rules. The learning algorithm of the neural network is the backpropagation algorithm. During the training, the network can add new fuzzy partitions due to the condition of the convergence, and then reconstructs itself to learn again. This method can generate a simple and effective rule set, and has a good convergence condition as well as a fast convergence speed