Abstract - Fuzzy logic is applied t

Abstract - Fuzzy logic is applied to the category discrimination problem related to identification of mammary lesions as benign or malignant. Results of other similar studies are reviewed. The current analysis expands the fuzzy logic approach by using the normal distribution function as set membership functions and using a genetic algorithm to optimize performance with the training partition. The approach is applicable to problems having arbitrarily large number of parameters. Two different data sets are examined. Data is portioned into a training set and validation set and each set is segregated into benign and malignant records. Values of mean and standard deviation are initially computed from the associated attributes and are different for the benign and malignant records. In one training method the standard deviations are adjusted to minimize overall error. In a second method a bias adjusts the importance of each membership function. Defuzzification is accomplished in three ways: modified averaging and OR process; comparison of multiplied fuzzy set values; and comparison of the multiplied squared set values. Results are compared with results obtained through statistical logistic regression.