Expert knowledge and supervised learning of rules: Application to Echinoderms

In this paper, we focus on expert knowledge incorporation in supervised learning tasks particularly decision rules. We aim to improve their quality, reduce their number and increase their prediction´s rate. The proposed method consists of initially applying Knowledge Discovery in Databases process (KDD) on a database relating to Echinoderms. It aims to improve then classification rules´ performance using background knowledge. This method is evaluated on a real domain area concerning Echinoderms. Experimental results are relevant; they generally improve performance and reduce prediction rules number.