A New Naïve Style Possibilistic Network Classifier
This paper proposes a new approach of classification under the possibilistic network (PN) framework with Tree Augmented Naïve Bayes Network classifier (TAN), which combines the advantages of both PN and TAN. The classifier is built from a training set where instances can be expressed by imperfect attributes and classes. A new operator, the possibilistic mean is designed to estimate the conditional possibility distributions of each attribute with imperfection, and the weight between two attributes given the class is determined by the conditioning specificity gain. Experiment has shown the efficiency of the new classifier in imperfect cases.
KeywordsTree augmented naïve bayes network Possibility theory Possibilistic classifier Imperfect cases
This work was supported by Doctor Subject Foundation of Heibei University of Science and Technology under Grant No. QD201051 and Scientific Research Plan of Hebei Education Department under Grant No. ZH2011243.
- 4.Borgelt C, Gebhardt J (1999) A naïve Bayes style possibilistic classifier. In: Proceedings of seventh European Congress on intelligent techniques and soft computing, Germany, pp 556–565Google Scholar
- 6.Zhang H, Jiang L, Su J (2005) Hidden naive Bayes. In: Proceedings of twentieth national conference on artificial intelligence (AAAI ’05), pp 919–924Google Scholar
- 7.Ben Amor N, Benferhat S, Dubois D et al (2003) Anytime propagation algorithm for min-based possibilistic graphs, Soft computing, A fusion of foundations. Methodologies and Applications, vol 8. Springer, Berlin, pp 150–161Google Scholar
- 8.Borgelt C (2000) Data mining with graphical models [D]. Otto-von-Guericke-University of Magdeburg, GermanyGoogle Scholar
- 9.Frank A, Asuncion A (2010) UCI machine learning repository. University of California, School of Information and Computer Science, Irvine [http://archive.ics.uci.edu/ml]