Ensemble Learning for Heart Disease Diagnosis: AVoting Classifier Approach

Abstract

Cardiovascular disease remains a serious public health problem internationally, responsible for a considerable number of fatalities. Early and correct detection of cardiovascular illness is crucial for optimal care and control of the condition. In this paper, we present an ensemble learning technique that includes voting classifiers to increase the reliability of cardiovascular disease diagnosis. We obtained a set of data from five cardiology databases, which included the Cleveland, Hungary, Switzerland, Long Beach VA and Statlog (Heart) datasets, which supplied us with a total of 1189 entries. We employed a feature engineering approach to extract relevant features from the dataset, enabling us to acquire vital information to enhance our model's performance. We trained and evaluated several machine learning algorithms, such as Random Forests, MLP, K-Nearest Neighbors, Extra Trees, XGBoost, Support Vector Machines, AdaBoost, Decision Trees, Linear Discriminant Analysis, and Gradient Boosting, and then incorporated these models using voting classifiers to produce more reliable and accurate models. Our findings reveal that the proposed ensemble learning process outperforms standalone models and conventional ensemble approaches, obtaining an accuracy rate of 91.4%. Our technique is likely to benefit clinicians in the early diagnosis of heart problems and improve patient outcomes. This work has major significance for the area of cardiology, indicating the possibility for machine learning approaches to boost both the reliability and accuracy of heart disease identification. The recommended ensemble learning technique may be adopted in hospitals to enhance patient care and eventually lessen the worldwide impact of cardiovascular disease. Further study is required to investigate the uses of predictive modeling in cardiology and other medical domains.