Performance Improvement of Heart Disease Prediction by Identifying Optimal Feature Sets Using Feature Selection Technique

Machine learning models are gaining popularity for its application on a wide range of medical diagnosis tasks. It is essential to make accurate predictions for such tasks as decisions have the potential to make a huge impact for the patient, even reducing fatality. In this paper, we have focused on increasing the accuracy as well as efficiency of prediction of heart disease using feature selection techniques. We have implemented a wrapper based sequential forward selection algorithm to identify optimum sets of features significant for the prediction of coronary artery disease. Our proposed feature selection technique was able to increase the accuracy of prediction by a maximum of 9.7% as opposed to not performing feature selection, and has the potential to be an effective and efficient tool in heart disease prediction.