Evaluating Binary Classifiers for Cardiovascular Disease Prediction: Enhancing Early Diagnostic Capabilities.

Area of Science:

• Medical Informatics
• Computational Biology
• Machine Learning in Healthcare

Background:

• Cardiovascular disease (CVD) is a leading global cause of mortality.
• Early detection of CVD is crucial for reducing complications and mortality rates.
• Machine learning (ML) offers promising tools for predicting CVD risk from patient data.

Purpose of the Study:

• To evaluate the effectiveness of seven binary classification algorithms for CVD risk prediction.
• To assess the impact of advanced preprocessing techniques on model performance.
• To identify the most accurate ML model for early CVD detection.

Main Methods:

• Applied seven classification algorithms: Random Forests, Logistic Regression, Naive Bayes, kNN, SVM, Gradient Boosting, and ANN.
• Utilized SMOTE-ENN for class imbalance and Grid Search Cross-Validation for hyperparameter optimization.
• Evaluated models using accuracy, precision, recall, F1-score, and ROC-AUC metrics.

Main Results:

• K-Nearest Neighbors (kNN) achieved the highest accuracy (99%) and ROC-AUC (0.99).
• kNN outperformed traditional models like Logistic Regression and Gradient Boosting.
• Addressing class imbalance and feature selection improved predictive model reliability.

Conclusions:

• kNN demonstrates significant potential as a reliable tool for early CVD prediction.
• Advanced ML techniques, when properly applied, enhance CVD risk assessment.
• This study provides a foundation for improved ML-based CVD prevention strategies.