MulO-AntMiner: A new ant colony algorithm for the multi-objective classification problem

Abstract

This paper presents a new ant-based algorithm for the multi-objective classification problem. The new algorithm called MulO-AntMiner (Multi-Objective Ant-Miner) is an improved version of the Ant-Miner algorithm, the first implementation of the ant colony algorithm for discovering classification rules. The fundamental principles in the proposed algorithm are almost similar to those in original Ant-Miner; even though, in our work there are two or more class attributes to be predicted. As a result, the consequent of a classification rule contains multiple predictions, each prediction involving a different class attribute. We have compared the performance of MulO-AntMiner with two other algorithms namely the C4.5 algorithm and the original Ant-Miner. © 2011 Springer-Verlag.