Tree Augmented Naïve possibilistic network classifier

Tree Augmented Naïve Bayes Network (TAN) classifier has shown excellent performance in Machine Learning and Data Mining in spite of the assumption of one- dependence of attributes. This paper proposes a new approach of classification under the possibilistic network (PN) framework with TAN, named tree augmented naïve possibilistic network classifier (TANPC), which combines the advantages of the PN and TAN. The classifier is built from a training set where instances can be expressed by imperfect attributes and classes. It is able to classify new instances those may have imperfect attributes.