The paper presents the rough set based methodologies for developing a predictive model for the occurrence of heart disease. The current study employs classical rough set theory (CRSA), which uses similarity relationships between the objects in the decision table to determine the inclusion of a certain condition in a decision class. The other variant of rough set methodology based on dominance relationship, Dominance-based Rough Set Analysis (DRSA) incorporates the monotonic relationship between the condition attributes and the desired variable to explain their respective decision classes. The data for this study is from Cleveland’s heart disease database of 303 heart disease patients taken from the open source online. Classification accuracy and explanatory power of the rough set based methods are compared with that of other machine learning languages and non-rule based methods. Non-rule-based methods performed marginally better than the rule based rough set methods, which cannot shadow the benefits of rough set methods provided by the provision of allowing a large amount of inconsistency in the decision rules.