Detection of Heart Disease Using Binary Classification Machine Learning Model

Abstract

Heart disease is a significant global health issue, causing millions of deaths annually. Despite advancements in medical technology, early and accurate diagnosis remains challenging. This study aims to detect heart diseases using binary classification machine learning models. The methodology employed a Heart Failure Prediction Dataset from Kaggle, with no issues of duplicates, missing data, outliers, or multicollinearity. Five machine learning models, including K-Neighbor Classifier, decision tree, support vector machine, random forest, and logistic regression, were trained and tested. The random forest model with hyper-parameters 'n_estimators': list (range (5,40,3)), 'max_features': ['log2', 'sqrt'] yielded the highest accuracy rate of 87.5%, precision rate of 90.4%, recall rate of 87.9%, f1_score of 89.1%, and auc_score of 93.6%. These results indicate that the random forest model has a notable capacity for accurate heart disease prediction, offering potential benefits such as reduced mortality rates and improved patient outcomes. Further research is recommended to establish standard data collection and analysis methods and to develop prediction models that consider the unique characteristics of diverse populations.