Arlei Silva,  Gisele L. Pappa,  Marcos A. Gonçalves,  Wagner Meira Jr.. 

This paper presents the first steps towards a genetic programming al­gorithm to evolve rule evaluation metrics for associative classifiers. The method allows the combination of characteristics found in a variety of metrics curren­tly used for rule evaluation, creating new and more robust evaluation functions. Experiments in the Reuters database showed that a simple associative classifi­cation algorithm combined with the evolved functions obtains better accuracy than the one using the confidence as an evaluation metric.

http://www.lbd.dcc.ufmg.br:8080/colecoes/waamd/2009/001.pdf

Processando consulta.