A comparison of bio-inspired metaheuristic approaches in classification tasks

Abstract

This paper presents a comparative analysis of three computational tools based in metaheuristics inspired by nature to perform an important data mining task. These tools are employed to generate classification rules from databases. The first one uses the Ant Colony metaphor that is one of the most recent nature-inspired metaheuristics. The second one employs the Artificial Immune System paradigm that is also a relatively new biologically-inspired paradigm. The third one employs a fuzzy genetic approach. The main motivation for applying those heuristics to data mining is that bio-inspired algorithms have shown to be robust search methods. In this work, basic concepts of the employed strategies are presented and significant aspects related to each approach are discussed. Some data sets from the UCI repository were employed to evaluate the performance of the tools. The comparative survey of the classification tasks is performed emphasizing the importance of discovering comprehensible and accurate knowledge.