An Efficient Pruning Approach for Class Association Rule Mining

Abstract

We propose an efficient pruning approach to build a faster classifier based on CARs (Class Association Rules). First, we develop a structure called LECR (Lattice of Equivalence Class Rules) and propose an algorithm for fast mining CARs. Second, we propose an algorithm to prune rules that are redundant in LECR. Experimental results show that our approach is more efficient than the one based on the ECR-tree (Equivalence Class Rules-tree) by Vo and Le in 2009. The rule sets generated by two approaches, ECR-tree and LECR, are the same, therefore the accuracy does not change.