Using knowledge-assisted discriminant analysis to generate new comparative terms

In this paper we present a method — knowledge-assisted discriminant analysis — to generate new comparative terms on symbolic and numeric attributes. The method has been implemented and tested on three real-world databases: mushroom classification, letter recognition, and liver disorder diagnosis. The experiment results show that combining AI and statistical techniques is an effective and efficient way to enhance a machine learning system’s concept description language in order to learn simple and comprehensible rules.