In this talk, Issam Salman will summarize his PhD thesis. First, he will discuss the implementation of a method for constructing a Tree-Augmented Naive Bayesian (TAN) model and a feature selection method called Selective TAN, which is specifically tailored for incomplete and unbalanced data. Second, an analysis of medical records of patients with acute myocardial infarction in both Syria and the Czech Republic will be presented. In this analysis, machine learning techniques were used to predict AMI mortality. Finally, a methodology for learning the structure of BN and Belief Noisy-Or models from incomplete data sets using Gaussian mixture will be presented.