Mining of generalized interesting classification rules with artificial chemical reaction optimization algorithm

Abstract

Classification rules mining is one of the most studied data mining problems. In this article, chemistry based Artificial Chemical Reaction Optimization Algorithm (ACROA) has been for the first time used for mining of generalized classification rules, which is a complex and no well researched generalized variant of classification rules mining where there is more than one goal attribute to be predicted. Furthermore, interestingness measure has been added by performing the adaptations to the algorithm in order to make the rules mined by the algorithm not only accurate and comprehensible but also interesting, surprising, and unexpected. Different rule sets within different databases satisfying different objectives have been flexibly mined by adapting the representation scheme and objective function. Performance of ACROA in classification rules mining within different real public databases have been compared that of genetic algorithm, particle swarm optimization algorithm, and ant colony optimization algorithm. It has shown that performance of ACROA within this special task of data mining is promising. ACROA can be an efficient solution method for different data mining tasks such as association rules mining, clustering rules mining, sequential pattern mining, and etc.