An Ensemble Machine Learning-Based Classification for Cardiovascular Disease Prediction Using PCA and SVM with Bagging

Abstract

Background: Cardiovascular diseases (CVD) remain the leading cause of mortality worldwide, highlighting the need for accurate and timely diagnosis. Machine learning has emerged as a promising tool for enhancing predictive accuracy in medical diagnosis.

Objective: This research aims to predict cardiovascular diseases using various machine learning algorithms applied to the Erbil Cardiovascular Health Dataset (Clinical data) from Mendeley. The objective is to improve prediction accuracy through ensemble learning and to extract the feature for better results and by calculating the feature importance.

Methods: The study involves data pre-processing using regression based filling by calculating mean, median for missing values, feature extraction using Principal Component Analysis (PCA), and for classification we are using the feature extracted dataset and applying Random Forest, Support Vector Machine (SVM), and Logistic Regression (LR). An ensemble method, bagging, was introduced to enhance model robustness and accuracy.

Results: The ensemble models demonstrated improved accuracy compared to independent models. Support Vector Machine with Bagging achieved 97% accuracy, Random Forest with Bagging reached 92%, and Logistic Regression with Bagging achieved 95%. Independent models without bagging showed lower accuracies: Support Vector Machine at 91.04%, Logistic Regression at 88.06%, and Random Forest at 83.58%. The effective of the ensemble method is evaluated using accuracy, precision, recall, log loss f1score.

Conclusion: The results indicate that machine learning models, especially ensemble methods, Support Vector Machine with Bagging achieved highest accuracy of 97% among the other algorithms and it can significantly enhance the early diagnosis and management of cardiovascular diseases, thereby improving patient care and outcomes.