ISLE - A Novel Immune-System Inspired Rule Extraction Algorithm

Authors:	Soroka Anthony, Cardiff University, United Kingdom Pham Duc, Cardiff University, United Kingdom
Topic:	5.1 Manufacturing Plant Control
Session:	Advanced Manufacturing Applications
Keywords:	Artificial Intelligence, learning algorithms, machine learning, knowledge-based systems, rules, rule-based systems

Abstract

This paper proposes a novel rule extraction algorithm whose extraction technique is inspired by the pattern recognition abilities of the immune system. This algorithm is based upon a form of immune network that facilitates the implementation of incremental rule extraction. The use of such a network also enables the algorithm to learn from training examples that do not have classes specified. The algorithm is tested against several example data sets and shows itself to be at least comparable to existing algorithms.