M. Eftekhari - Department of Computer Science and EngineeringSchool of EngineeringShiraz University
GH. Yaghoobi
S.D Katebi

The major problem of rule based Fuzzy Learning Classifier Systems (FLCS) is to design a good measure of rule selection. The fuzzy extension of confidence and support are two wellknown criteria that have been frequently used for evaluating fuzzy rules of a FLCS. In recent studies, the better classification accuracy is reported when composite measures of confidence and support have been considered for rule selection. Moreover, An Information Retrieval (IR) system can be evaluated for successfully retrieving objects by measures namely precision, recall and fallout. In this paper, the well-known Lp-Norm is employed for combining precision, recall and fallout to a scalar value. The resulted classification accuracy rates show the effectiveness of Lp-Norm for combining the IR measures.

Iterative rule learning, Rule evaluation criteria,Fuzzy classification, Information retrieval