An active TAN classifier based on vote entropy-maximum entropy of QBC

Abstract

Tree-Augmented Naive Bayes (TAN) is a state-of-the-art extension of the naive Bayes, which outperforms naive Bayes, yet at the same time maintains the computational simplicity and robustness that are characteristic of naive Bayes. But TAN classifier was built by the conventional passive learning. The available training samples with actual classes are not enough for passive learning method for modeling TAN classifier in practice. The Query-by-Committee (QBC) method of active learning can examine unlabelled examples and selects only those that are most informative for labeling. It aims at using few labeled training examples to build efficient classifier. In this paper, an active TAN classifier algorithm based on vote entropy-maximum entropy of QBC is presented to solve the problem of building TAN classifier from unlabelled samples. Experimental results show that the proposed algorithm can reach the same accuracy as passive learning with few labeled training examples.