A New Hierarchical Redundancy Eliminated Tree Augmented Naive Bayes Classifier for Coping with Gene Ontology-based Features.

The Tree Augmented Na¨ive Bayes classifieris a type of probabilistic graphical model thatcan represent some feature dependencies. Inthis work, we propose a Hierarchical RedundancyEliminated Tree Augmented Na¨ive Bayes(HRE–TAN) algorithm, which considers removingthe hierarchical redundancy during the classifierlearning process, when coping with data containinghierarchically structured features. Theexperiments showed that HRE–TAN obtains significantlybetter predictive performance than theconventional Tree Augmented Na¨ive Bayes classifier,and enhanced the robustness against imbalancedclass distributions, in aging-related genedatasets with Gene Ontology terms used as features.