The self-generating fuzzy algorithm with singleton output type for multi-input fuzzy variables
Kwang‐Yong Kim, Young-Kyu Yang · 1999
In this paper, we suggest a new self-generating fuzzy system without the use of other information processing techniques like-neural networks or genetic algorithms. The constraint of this system is that the conclusion part of fuzzy rules consists of the singleton output. This system can suppress notably increasing fuzzy rules and minimize the number of membership functions, as compared to the pure gradient type method suggested by Araki's method (1991). Several numerical examples show that the proposal method has the least number of fuzzy rules and the least number of membership functions to satisfy any desired accuracy in comparison with the old method (Araki's method). We also show a relation between experimental data and rules in comparison with the old method. In conclusion, our suggested method is efficient for multi-input fuzzy variables.