This paper describes the use of a Hybrid Fuzzy-Genetic Programming system to discover patterns in large databases. It does this by evolving a series of variable-length fuzzy rules which generalise from a training set of labelled classes. Numerous novel techniques, including the use of genotypes in Genetic Programming, two new genetic crossover operators, and the processes of Modal Evolution, Modal Reevolution and Nested Evolutionary Search are described. Experimental results show that the system is able to classify data from the Wisconsin Breast Cancer database correctly 95percent of the time.
展开▼
