Heart Disease Prediction Using K-Nearest Neighbor

Abstract

heart disease is a Non-Communicable Disease (NCDs) on cardiovascular system. It has long-term impact that can turn to death, but it has no symptoms that make it hard to recognize. The heart disease prediction believed to detect the presence of heart disease. Recent developments use various approach with machine learning, one of those is K-Nearest Neighbor (KNN). It will become a computer aid system to diagnose the heart disease. A drawback of using KNN is still achieve low accuracy that make the prediction is not helpful. The aim of this study is to improve the performance of heart disease prediction using standardization dataset and feature selection for KNN. We perform method to prepare data such as KNN imputation and standardization, also feature selection with SelectKBest. The performance is validated by 10-Fold Cross Validation. This paper present better precision of 95.50%, recall of 84.16%, and accuracy 89.28% for 0.59 ms. This study provides alternative method for processed data on heart disease prediction. We compare latest study to discover better solution in this field. The result will evolve the knowledge of machine learning in medical purpose.