This paper studies data mining techniques used in medical diagnosis, particularly for predicting chance of survival of a patient after undergoing thoracic surgery. We discuss models built using decision trees, naive Bayes and support vector machines and explore suitability of each of the algorithms to perform on such data.