Investigating Differential Evolution based rule discovery in learning classifier systems

Abstract

Differential Evolution is an efficient and powerful population-based stochastic search technique that has been applied mainly to optimization problems over continuous spaces. Despite its potential only a few researchers have recently explored its use in the machine learning domain, specifically for clustering problems. In this paper, we investigate the use of differential evolution for classification rule discovery using a learning classifier systems framework. Learning classifier systems are genetics-based machine learning techniques that have recently shown a high degree of competence on a variety of data mining problems. They use a niched genetic algorithm for rule discovery and generalization. Stalling of genetic search when dealing with high dimensional real-valued classification problems is a common problem in learning classifier systems. A new rule discovery component based on differential evolution is proposed in this paper to improve learning classifier systems' search capabilities. The experimental results indicate that the proposed approach increases the classification accuracy and convergence speed of the system.