Mining fuzzy classification rules using an artificial immune system with boosting

Abstract

In this study, a classification model including fuzzy system, artificial immune system (AIS), and boosting is proposed. The model is mainly focused on the clonal selection principle of biological immune system and evolves a population of antibodies, where each antibody represents the antecedent of a fuzzy classification rule while each antigen represents an instance. The fuzzy classification rules are mined in an incremental fashion, in that the AIS optimizes one rule at a time. The boosting mechanism that is used to increase the accuracy rates of the rules reduces the weight of training instances that are correctly classified by the new rule. Whenever AIS mines a rule, this rule is added to the mined rule list and mining of next rule focuses on rules that account for the currently uncovered or misclassified instances. The results obtained by proposed approach are analyzed with respect to predictive accuracy and simplicity and compared with C4.5Rules.