Use of Attribute Selection Criteria in Decision Trees in Uncertain Domains
Bruno Crémilleux, Christian P. Robert · Advances in fuzzy systems · 2000
There are many attribute selection criteria involved in the induction of decision trees. We present criteria derived from an impurity measure in a unified framework and we use the same background to describe ORT criterion [12]. We compare the theoretical backgrounds of C.M. criteria with those of the ORT criterion and we illustrate their differences. We set lazy decision trees with regards to these families of criteria, and we debate their use with uncertain data sets. Finally, we clarify the choice of these approaches and we separate their domains of application.