Tree-Augmented Naive Bayes

The Tree Augmented Naive Bayes also relaxes the strong conditional independence assumption of the Naive Bayes network. The naive structure is augmented with a Maximum Weight Spanning Tree. The predictive performance of this algorithm is better than a Naive Bayes network, but not as good as an Augmented Naive Bayes network. However, with the tree constraint in learning, the Tree Augmented Naive Bayes algorithm is much faster than the Augmented Naive Bayes algorithm.