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.