Three discretization methods for rule induction

We discuss problems associated with induction of decision rules from data w ith numerical attributes. Real-life data frequently contain numerical attri butes. Rule induction from numerical data requires an additional step calle d discretization. In this step numerical values are converted into interval s. Most existing discretization methods are used before rule induction, as a part of data preprocessing. Some methods discretize numerical attributes while learning decision rules. We compare the classification accuracy of a discretization method based on conditional entropy, applied before rule ind uction, with two newly proposed methods, incorporated directly into the rul e induction algorithm LEM2, where discretization and rule induction are per formed at the same time. In all three approaches the same system is used fo r classification of new, unseen data. As a result, we conclude that an erro r rate for all three methods does not show significant difference, however, rules induced by the two new methods are simpler and stronger. (C) 2001 Jo hn Wiley & Sons, Inc.