Comparing Bayesian Network Classifiers

In this paper, we empirically evaluate algorithms for learning four Bayesian network (BN) classifiers: Naïve-Bayes, tree augmented Naïve-Bayes (TANs), BN augmented NaïveBayes (BANs) and general BNs (GBNs), where the GBNs and BANs are learned using two variants of a conditional independence based BN-learning algorithm. Experimental results show the GBNs and BANs learned using the proposing learning algorithms are competitive with (or superior to) the best classifiers based on both Bayesian networks and other formalisms, and that the computational time for learning and using these classifiers is relatively small. These results argue that BN classifiers deserve more attention in machine learning and data mining communities. 1 INTRODUCTION Classification is a fundamental task in fault diagnosis, pattern recognition and forecasting. In general, a classifier is a function that chooses a class label (from a group of predefined labels) for instances described by a set of features (attributes..