Efficient Pruning Methods for Obtaining Compact Associative Classifiers with Enhanced Classification Accuracy Rate

Abstract

The integrated approach involving supervised classification and association rule mining for development of classification model is becoming a promising strategy for building compact and accurate classifiers. As discussed in literature, in large databases the association rule mining techniques may produce large rule sets, this paper attempts to propose and implement several pruning methods to overcome the problems underlying the Associative Classification approach. The aim is to efficiently utilize the rules produced for classification in compact form and represent a rule set to be the part of classifier model with maximum data coverage and enhanced accuracy. This paper also presents experimental evaluation of the pruning methods on various datasets taken from UCI machine learning repository with the consideration of earlier approaches.