Associative Classification Based on Artificial Immune System

Associative classification algorithms which are based on association rules have performed well compared with other classification approaches. However a fundamental limitation with these classification algorithms is that the search space of candidate rules is very large and the processes of rule discovery and rule selection are conducted separately. This paper proposes an approach called ARMBIS, which is based on the natural immune principle, for searching associative rules. The proposed algorithm has the capability of dealing with complex search space of association rules while still ensuring that the resultant set of association rules is appropriate for associative classification. The performance evaluation results have shown that the proposed algorithm has achieved good runtime and accuracy performance in comparison with conventional associative classification algorithms.