Feature Selection with Combination Classifier use Rules-Based Data Mining for Diagnosis of Coronary Heart Disease

Coronary heart disease often causes of deaths on human. This disease occurs when there is atherosclerosis (fat deposits) that block the flow of blood to the heart muscle in the coronary arteries. The gold standard method that doctors refer to diagnose coronary heart disease is coronary angiography. However, this method is invasive, high risk, expensive and sometimes the diagnosis result is not accurate. The purpose of this research is to perform diagnosis of coronary heart disease based on computer. Diagnosis of coronary heart disease based on computer will be done with data mining methods rules, based on Cleveland dataset. In this research, feature selection method based on computer and medical expert used are Variable Precision Rough Set (VPRS) and Motivated Feature Selection (MTF). The classification for diagnosis used data mining methods based on rules are VPRS and Repeated Incremental Pruning Error Reduction (RIPPER). The methods are chosen to observe the simplest pattern of features and rules knowledge from big data, imprecise and ambiguous data. The proposed method is evaluated on Cleveland coronary heart disease dataset taken from the UCI repository. The feature selection VPRS and the combination classifier VPRS dan RIPPER obtains the best evaluation result with accuravy achieved of 88,88889%. While the accuracy of VPRS and RIPPER have the same values is 84,84848%. It indicated that the proposed combination methods successfully classifies coronary heart disease dataset dan has a potential to be implemented in the development of a computerized coronary heart disease diagnosis system.