Decision tree induction methods using an entropy criterion

Paul-Benoît Perche , Denis Pomorski (1) , Marcel Staroswiecki (1)

1 LAGIS - Laboratoire d'Automatique, Génie Informatique et Signal

In this contribution, we propose an approach in order to construct a set of rules from observed data, which allows to obtain a qualitatif model of the system : an atomic model, based on the notion of filtering of the contingency table, is generalized via decision tree induction methods using a conditional entropy criterion. We propose two approaches of the multi-variable knowledge problem (as part of the decision tree induction methods).

Mots clés: decision tree; entropy

Domaines: Théorie de l'information [cs.IT]

Paul-Benoît Perche, Denis Pomorski, Marcel Staroswiecki. Decision tree induction methods using an entropy criterion. Second International Conference on Industrial Automation, Jun 1995, Nancy, France. ⟨hal-01509942⟩