AUTOMATED RULE-BASE CREATION via CLIPS-INDUCE

Many CLIPS rule-bases contain one or more rule groups that perform classi cation. In this paper we describe CLIPS-Induce, an automated system for the creation of a CLIPS classi cation rule-base from a set of test cases. CLIPS-Induce consists of two components, a decision tree induction component and a CLIPS production extraction component. ID3 [1], a popular decision tree induction algorithm, is used to induce a decision tree from the test cases. CLIPS production extraction is accomplished through a top-down traversal of the decision tree. Nodes of the tree are used to construct query rules, and branches of the tree are used to construct classi cation rules. The learned CLIPS productions may easily be incorporated into a large CLIPS system that perform tasks such as accessing a database or displaying information.