Detection of Coronary Heart Disease Using Modified K-NN Method with Recursive Feature Elimination

Abstract

Heart disease has infected many people in the world. One of the most common and deadly heart diseases is coronary artery disease (CAD). Several parameters can diagnose a patient who is positive for coronary artery disease (CAD). These parameters based on demographic, symptom and examination, ECG, and laboratory and echo features. For prevention needs to be done by early detection of patients who have the potential to have CAD. One way to do early detection is by building a system for making predictions. In this study, researchers used the KNN method by developing the weight of each class to increase accuracy. Before entering the KNN, the data will perform feature selection using SVM-RFE to find the ideal features and speed up computing time. The results of the KNN without feature selection are 82.65% of accuracy, the KNN with feature selection achieves 86.33% of accuracy, the Weighted KNN without feature selection achieves 83.45% of accuracy, and the Weighted KNN with feature selection achieves 90.88% of accuracy. The results prove the effectiveness of the Weighted KNN with feature selection.