Classification of Cardiovascular Disease Patients Using Feature Selection and K-Nearest Neighbor

Cardiovascular disease is a non-communicable disease that is the leading cause of death in the world. By understanding the factors causing the disease, data mining can be used as an early prevention so as not to worsen the sufferers of the disease. This study aims to classify patients with cardiovascular disease using the KNN algorithm. In addition, to improve the performance of the KNN model created, feature selection namely forward selection and backward elimination is used to select attributes, and the grid operator is used to find the appropriate K value for KNN modeling. Evaluation of the KNN model was carried out on the Cardiovascular Diseases Dataset using 10 -fold cross-validation. The backward elimination operator has an accuracy of $\mathbf{7 2. 9 2 \%}$, recall of $\mathbf{6 7. 3 1 \%}$, and f1score of $\mathbf{7 1. 1 0 \%}$ which is superior to the forward elimination operator with an accuracy of $\mathbf{7 2. 8 0 \%}$, recall of $\mathbf{6 6. 6 3 \%}$, and f1score of $70.81 \%$. In addition, the forward selection operator has a precision of $\mathbf{7 5. 5 4 \%}$ and an average execution time of $\mathbf{5 2: 5 0}$ which is superior to the backward elimination operator with a precision of $75.35 \%$ and an execution time of $1: 27: 41$. This study shows the potential that the KNN algorithm using the backward elimination is able to classify patients suffering from cardiovascular disease more accurately and efficiently.