Conclusions
It is clear that fuzzy logic employing a exponential membership function, weighted averaging, and logical OR operation is an effective tool in data analysis involving discrimination where many parameters are involved. It performs as well as results reported by other researchers looking at similar problems with different fuzzy methods and hybrid fuzzy methods. Because membership functions may be adjusted individually the usual method of dividing data into a training partition and a validation partition is appropriate. Rather extensive adjustment (“learning”) during the training stage improves performance. This may include use of a genetic algorithm. Results observed here must be classified as anecdotal because only two data sets were used. However, from these observations it appears that fuzzy logic is comparable but does not outperform usual data analysis methods such as logistic regression. In instances where image analysis is a principal feature (as reported in other papers noted earlier) the situation does not lend itself to standard statistical regression methods.