Performance Improvement with Decision Tree in Predicting Heart Failure

Abstract

Cardiovascular diseases is a general term given to the group of diseases that includes heart failure, heart attack, stroke. They are quite dangerous for human health. Various studies have been conducted in the literature to predict the survival of patients with heart failure. In this study, user-defined parameters of three different machine learning methods (logistic regression-LR, K nearest neighbor-KNN, and decision tree-DT) used in existing studies are optimized to make predictions with higher accuracy. In terms of objectivity and reliability of the experimental results, k-fold cross validation technique is applied. As a result, the performance results of this study are observed to be 10% and 3% higher than the literature in the DT and KNN algorithms, respectively. In particular, the proposed KNN method has shown that it can guide physicians in the decision-making process.