Output details

A New Sequential Covering Strategy for Inducing Classification Rules with Ant Colony Algorithms

<24> This paper proposes a significant improvement in the sequential covering strategy used in ant colony classification algorithms. Since the first version of the Ant-Miner classification algorithm (2002), these algorithms have employed the strategy to create a list of rules one rule at a time until all training examples are covered. This paper discusses the drawbacks of this strategy and proposes a new sequential covering strategy addressing its limitations. The results comparing a novel algorithm implementing the new strategy against state-of-the-art rule induction algorithms show that the new strategy leads to statistically significant improvements in terms of predictive accuracy.