Ranking with fuzzy decision trees
Xuan Tuan Le, Christophe Marsala · 2013
Ranking objects with decision trees has recently received much attention of researchers, with prospect of utilizing significant advantages of the tree model. This paper checks the ranking ability of fuzzy decision trees in bipartite ranking setting. The main reason for the introduction of fuzzy decision trees in classification is the presence of imprecise and imperfect data which is unhandled by classical trees. In this study, we introduce another advantage of fuzzy decision trees: their ability in performing instance-ranking based on class membership degrees associated with each instance. The ranking abilities of other methods using classical decision tree are also examined. Experiments show that the ranking performance of fuzzy decision trees is better than others on clean datasets and outperforms them on noisy datasets.