Improving Classification Performance by Combining Multiple TANClassifiers

Boosting is an effective classifier combination method, which can improve classification performance of an unstable learning algorithm. But it does not have much more improvements on a stable learning algorithm. TAN, Tree-Augmented Naive Bayes, is a tree-like Bayesian network. The standard TAN learning algorithm generates a stable TAN classifier, which is difficult to improve its accuracy by boosting technique. In this paper, a new TAN learning algorithm called GTAN and a TAN classifier combination method called Boosting-MultiTAN are presented. Through comparisons of this TAN classifier combination method with the standard TAN classifier in the experiments, the Boosting-MultiTAN shows higher classification accuracy than the standard TAN classifier on the most data sets.