A Machine Learning Approach to Heart Attack Risk Assessment Using Feature-Engineered Clinical Data

Cardiovascular diseases (CVDs) remain one of the leading causes of deaths throughout the world and that essentially emphasizes the necessity of developing accurate risk prediction models. This paper presents a feature engineered ensemble learning approach to heart attack risk prediction based on a wide range of clinical and lifestyles factors. The dataset, which contained 8760 samples and 28 features, was optimized with the immense feature engineering, which included creating new features such as Systolic BP (Blood Pressure), Diastolic BP, BP Ratio and Cholesterol to Triglycerides. Various machine learning algorithms such as Logistic Regression, KNN, Decision Tree, Naive Bayes, SVM, GBM, XGBoost, and MLP were compared to each other with respect to accuracy, precision, recall and F1-score. The result of the model demonstrated that the traditional models performed moderately well with Logistic Regression having a recall value of 0.99 and poor generalizability while ensemble models such as GBM and XGBoost performed better in terms of stability and accuracy. The Proposed Feature Engineering + Ensemble Model using Random Forest and XGBoost proved better than all other models with an accuracy of 0.92 and F1 score of 0.84 and hence establishes the superiority of the hybrid ensemble learning method to improve the accuracy of the diagnosis.