SWARM INTELLIGENCE FOR RULE DISCOVERY IN DATA MINING

This paper aims to discuss Swarm Intelligence approaches for Rule Discovery in Data Mining. The first approach is a new rule learning algorithm based on Particle Swarm Optimization (PSO) and that uses a Multiobjective technique to conceive a complete novel approach to induce classifiers, called MOPSO-N. In this approach the properties of the rules can be expressed in different objectives and then the algorithm finds these rules in an unique run by exploring Pareto dominance concepts. The second approach, called PSO/ACO2 algorithm, uses a hybrid technique combining Particle Swarm Optimization and Ant Colony Optimization. Both approaches directly deal with continuous and nominal attribute values, a feature that current bioinspired rule induction algorithms lack. In this work, an experiment is performed to evaluated both approaches by comparing the performance of the induced classifiers.