Cost sensitive discretization of numeric attributes

Abstract

Many algorithms in decision tree learning are not designed to handle numeric valued attributes very well. Therefore, discretization of the continuous feature space has to be carried out. In this article we introduce the concept of cost sensitive discretization as a preprocessing step to induction of a classifier and as an elaboration of the error-based discretization method to obtain an optimal multi-interval splitting for each numeric attribute. A transparant description of the method and steps involved in cost sensitive discretization is given. We also evaluate its performance against two other well known methods, i.e. entropy-based discretization and pure error-based discretization on a real life financial dataset. From the algoritmic point of view, we show that an important deficiency from error-based discretization methods can be solved by introducing costs. From the application point of view, we discovered that using a discretization method is recommended. To conclude, we use ROC-curves to illustrate that under particular conditions cost-based discretization may be optimal.