A Simple Technique for Generation and Minimization of Fuzzy Rules

The popular techniques of rule generation from numerical data such as neural networks, genetic algorithms, and fuzzy clustering are suitable when the available data pairs are large. In case of limited available data sets, a new approach for rule generation and minimization has been proposed in the present paper. Initial rules for each data pairs are generated and conflicting rules are merged based on their degree of soundness. The minimization technique for membership functions differs from others in the sense that the two or more membership functions are not merged but replaced by a new membership function whose minimum and maximum ranges are the minimum value of the first and maximum of the last membership function and bisection point of the two or more is the peak of new membership function. The proposed scheme has been applied to predict one of the important effects (i.e. annoyance) of noise pollution on human beings. The data is based on the reports of Environmental Protection Agency (EPA) published in 1977 for the surveys conducted in several metropolitan cities of USA