Performance Evaluation of Machine Learning Model in Predicting Heart Disease Prevalent

Abstract

Heart disease stands as the foremost contributor to global mortality in the present era. While a range of medical tests aids in diagnosing various manifestations of heart disease, anticipating its occurrence without resorting to these tests presents a formidable challenge. Machine learning has revealed a potential path for comprehending extensive medical datasets, unearthing concealed insights that elude human observation. Random Forest, Logistic Regression, XG Boost, k-nearest neighbor, and Gradient boosting was used to predict instances of heart diseases based on age and gender in this study. Regardless of the dataset's generalization or gender-specific characteristics, we observed that Random Forest, XG Boost, and K-Nearest Neighbors models exhibit high robustness and minimal sensitivity to hyperparameters. While the prediction accuracy remains consistent among Random Forest, Xgboost, and K-Nearest Neighbors, the crucial features responsible for achieving that accuracy vary among them. This observation indicates that it is important to test different machine learning algorithms when the interest is in prediction and understanding the most important feature(s). Keywords: Machine Learning, Heart Disease, Diagnosis, Predictions, Accuracy CISDI Journal Reference Format Abiola O. A. Adebisi R. O. Adeyemo A. B. (2024): Performance Evaluation of Machine Learning Model in Predicting Heart Disease Prevalent. Computing, Information Systems, Development Informatics & Allied Research Journal. Vol 15 No 1, Pp 63-72.dx.doi.org/10.22624/AIMS/CISDI/V15N1P5. Available online at www.isteams.net/cisdijournal