A proposed technique for predicting heart disease using machine learning algorithms and an explainable AI method.

Area of Science:

• Medical data analysis
• Machine learning applications in healthcare
• Cardiovascular disease research

Background:

• Accurate heart disease prediction is crucial for early intervention and reducing mortality.
• Current methods are limited by the inability of healthcare professionals to provide constant patient supervision.
• Machine learning (ML) offers a data-driven approach to enhance prediction and decision-making in heart disease detection.

Purpose of the Study:

• To develop an accurate ML algorithm for early heart disease prediction using diverse feature selection strategies.
• To identify the most effective ML algorithm and feature subset for predicting heart disease.
• To create a mobile application for instant heart disease risk assessment based on symptoms.

Main Methods:

• Feature selection using chi-square, ANOVA, and mutual information methods (SF-1, SF-2, SF-3).
• Evaluation of ten ML algorithms including SVM, XGBoost, and Random Forests on private and public datasets.
• Application of Synthetic Minority Oversampling Technique (SMOTE) for data balancing and SHAP for explainable AI.

Main Results:

• The XGBoost algorithm, using the SF-2 feature subset on combined datasets, achieved optimal performance.
• Key performance metrics included 97.57% accuracy, 96.61% sensitivity, 90.48% specificity, 95.00% precision, 92.68% F1 score, and 98% AUC.
• An explainable AI method using SHAP was developed to interpret model predictions.

Conclusions:

• The proposed ML approach enables rapid and cost-effective early-stage heart disease identification for healthcare providers.
• The developed mobile app provides instant heart disease risk prediction based on user-input symptoms.
• The study highlights the potential of advanced ML techniques, like XGBoost and SHAP, in improving cardiovascular diagnostics.