Prediction of Heart Disease Using Machine Learning Algorithms

Abstract

Disease diagnosis aids clinicians in making appropriate therapy suggestions for patients. Heart disease is one of the most frequent diseases today, and early detection of the condition is critical for many health care providers in order to protect their patients and save lives. In this sector, one of the solutions for detecting disease-related symptoms is to utilize machine learning. Medical datasets have made machine-learning algorithms more intuitive. Machine learning algorithms offer methods for identifying datasets of features and their interactions. Various machine-learning techniques are analyzed on heart disease datasets in attempts to accurately identify and or forecast cases of heart disease.  In this study, seven parameter tuned machine learning algorithms were used to integrate and verify 14 variables of 920 patient data from Cleveland, Hungary, Switzerland, and Long Beach. K-Nearest Neighbor, Random Forest, Gaussian Naive Bayes, Logistic Regression, Support Vector Machine, Adaboost, and Gradient Boosting methods are among the hyperparameters tailored algorithms used. In order to compare the performance of the models, five performance metrics were used: accuracy, precision, sensitivity, F1 score, and Area Under the Curve. The findings were compared to machine learning methods that used the same-pooled datasets. It was observed that using different classification algorithms for the classification of the combined dataset and tuning the parameters of these algorithms produced very promising results in terms of classification accuracy for the Adaboost, Support Vector Machine, Naive Bayes, and Gradient Boost classifiers, with classification accuracies of 85.98%, 85.87%, 84.56%, and 84.39%, respectively.