Learning Restricted Bayesian Network Classifiers with Mixed Non-i.i.d. Sampling

Generally, numerous data may increase the statistical power. However, many algorithms in data mining community only focus on small samples. This is because when the sample size increases, the data set is not necessarily identically distributed in spite of being generated by some common data generating mechanism. In this paper, we realize restricted Bayesian network classifiers are robust even when training data set is non-i.i.d. sampling. Empirical studies show that these algorithms performs as well as others which combine independent experimental results by some statistical methods.