Decision tree classifiers for evidential attribute values and class labels

Decision trees are well-known machine learning techniques for solving complex classification problems. Despite their great success, the standard decision tree algorithms do not have the ability to process imperfect knowledge, meaning uncertain, imprecise and incomplete data. In this paper, we develop new decision tree approaches to cope with data that have uncertain attribute values and class labels. More concretely, we tackle the case where the uncertainty is represented and managed through the evidence theory. (C) 2018 Published by Elsevier B.V.