An Introduction to the Robust Bayesian Classifier

Bayesian supervised classifiers are one of the most promising data mining techniques for large scale applications. When the database is complete, they provide an efficient and scalable solution to classification problems. When some data are missing in the training set, methods exist to learn these classifiers, albeit less efficiently, under the assumption that data are missing at random. This paper describes the implementation of RoC, a Bayesian classifier able handle incomplete databases with no assumption about the pattern of missing data.

1. Knowledge Media Institute, The Open University

2. Department of Statistics, The Open University