Comprehensive Comparative Analysis of Computational Intelligence Models for Heart Disease Prediction

Abstract

Heart disease is a major global public health concern, highlighting the urgent need for precise prediction models to enhance prevention and early detection. This research looks at how well different computer models can predict heart disease. It mainly looks at Random Forests (RF), Logistic Regression (LR), Convolutional Neural Networks (CNNs), Decision Trees (DTs), and Artificial Neural Networks (ANNs). According to our data, ANNs were the most accurate model for prediction, achieving the highest accuracy (87.0%) with balanced recall (0.84), strong precision (0.91), and F1-score (0.87). CNNs displayed solid overall performance with an accuracy of 84.0%, precision and recall both at 0.85, and an F1 score of 0.85, suggesting their effectiveness in learning spatial hierarchies. Despite logistic regression attaining an accuracy of 82.0%, its precision (0.89) and recall (0.79) indicate a compromise between accurately detecting genuine positives and reducing false positives. Both Decision Trees and Random Forests achieved perfect precision, recall, and F1-scores (all 1.00), though Decision Trees had a lower accuracy of 74.0%, potentially indicating overfitting. This comparative analysis highlights the advantages and disadvantages of each model, offering insights into their usefulness for predicting heart disease. By incorporating essential metrics such as accuracy, precision, recall, and F1-score, this research contributes to creating more precise and dependable diagnostic tools, setting the stage for improved prevention and early intervention tactics for heart disease.