A genetic classifier tool

Knowledge discovery is the most desirable end product of an enterprise information system. Research from different areas recognizes that a new generation of intelligent tools for automated data mining is needed to deal with large databases. In this sense, induction based learning systems have emerged as a promising approach. This paper describes an induction-based classifier tool. The tool employs a genetic algorithm using the Michigan approach to find rules, is able to process discrete and continuous attributes and also is domain-independent. Implementation details are explained, including some optimizations, data structures and genetic operators. Some optimizations include the use of phenotypic sharing (with linear complexity) to direct the search. The results of accuracy are compared with 33 other algorithms in 32 datasets. The difference of accuracy is not statistically significant at the 10% level when compared with the best of the other 33 algorithms