Expert-assisted classification rules extraction algorithm

Abstract

Machine learning algorithms nowadays are important and well-accepted tools which help in demanding and ever-more challenging data analysis in many fields. In this paper, we study an approach to machine learning and knowledge discovery, where a learning algorithm uses experts' domain knowledge to induce solutions, and experts use the algorithm and its solutions to enhance their "information processing strength". An adaptation of evolutionary method AREX for automatic extraction of rules is presented that is based on the evolutionary induction of decision trees and automatic programming. The method is evaluated in a case study on a medical dataset. The obtained results are assessed to evaluate the strength and potential of the proposed classification rules extraction algorithm. © 2010 Springer-Verlag.