IBP-Laforia
1994/21:
Rapport de Recherche Laforia /
Laforia research reports
20 pages - Novembre/November 1994 -
French document.
PostScript : 63 Ko /Kb
Titre / Title: Arbres de décision et Sous-ensembles flous
Abstract : In this paper, we highlight the problem of numerical attributes in decision trees, particularly in the ID3 family decision trees.
We propose to discretize these attributes with an entropy measure during the construction of the tree.
We incorporate some elements of the fuzzy sets theory to classify an object with this tree. A path from the root of the tree to a leaf is associated with a value of a membership function. The degrees of all the paths are agregated to obtain the object's class. Various agregation fuzzy operators (such as t-norms and t-conorms) are used.
Publications internes Laforia 1994 / Laforia research reports 1994