Various techniques have been developed to improve
the performance of naive Bayes-many of them aimed
at reducing the 'naivete' of the algorithm-while still
retaining the desirable aspects of simplicity and computational
efficiency. Zheng and Webb (2000) provide
an excellent overview of work in this area. Most
existing techniques involve restricted sub-classes of
Bayesian networks, combine attribute selection with
naive Bayes, or incorporate naive Bayes models into
another type of classifier (such as a decision tree).