Naive Bayes is one of the most efficient and effective inductive learning algorithms for machine learning and data mining. But the conditional independence assumption on which it is based, is rarely true in real-world applications. Researchers extended naive Bayes to represent dependence explicitly, and proposed related learning algorithms based on dependence. In this paper, we argue that, from the classification point of view, dependence distribution plays a crucial role, rather than dependence. We propose a novel explanation on the superb classification performance of naive Bayes. To verify our idea, we design and conduct experiments by extending the ChowLiu algorithm to use the dependence distribution to construct TAN, instead of using mutual information that only reflects the dependencies among attributes. The empirical results provide evidences to support our new explanation.
[1]
Pedro M. Domingos,et al.
Beyond Independence: Conditions for the Optimality of the Simple Bayesian Classifier
,
1996,
ICML.
[2]
Pedro M. Domingos,et al.
On the Optimality of the Simple Bayesian Classifier under Zero-One Loss
,
1997,
Machine Learning.
[3]
Nir Friedman,et al.
Bayesian Network Classifiers
,
1997,
Machine Learning.
[4]
Catherine Blake,et al.
UCI Repository of machine learning databases
,
1998
.
[5]
C. N. Liu,et al.
Approximating discrete probability distributions with dependence trees
,
1968,
IEEE Trans. Inf. Theory.
[6]
Christopher J. Merz,et al.
UCI Repository of Machine Learning Databases
,
1996
.