Optimizing Cardiovascular Disease Prediction with a Hybrid Gradient Descent Adaptive Algorithm and Random Forest Classifier

Abstract

Cardiovascular diseases (CVDs) constitute a major global health burden, thereby emphasizing the critical need for the development of robust and accurate predictive models to ensure early detection and prompt clinical intervention. This study introduces a novel Hybrid Gradient Descent Adaptive Algorithm and Random Forest (Hybrid GD-AA-RF) model, which combines the strengths of AdaGrad and AdamW optimizers for effective feature selection and integrates them with a Random Forest (RF) classifier to enhance predictive accuracy. The proposed model employs Z-score normalization during preprocessing to standardize features and ensure consistency. Feature selection leverages Gaussian-based differential entropy for information gain, while the hybrid optimization technique combines AdaGrad’s adaptive learning rates with AdamW’s weight decay regularization to prioritize critical but infrequent features, minimizing overfitting and improving generalization.

The RF classifier dynamically adjusts Tuning parameters, like the number of estimators and maximum tree depth, optimizing its performance on high-dimensional medical datasets. The Hybrid GD-AA-RF model was evaluated on a combined heart disease dataset with 12 attributes and 1190 records. Comparative analysis with cutting-edge models demonstrated superior performance in the considered metrics, owing to the balanced attribute selection and categorization capabilities. The hybrid optimization approach avoids the complexity of metaheuristic algorithms while ensuring efficient computation and enhanced interpretability.

This model’s robust generalization across diverse data distributions and populations highlights its scalability for real-world healthcare applications. Its ability to prioritize impactful predictors can assist clinicians in identifying critical risk factors, enabling early diagnosis and improved patient outcomes. Marks a significant advancement in machine learning-based diagnostics, providing a dependable and precise tool for predicting cardiovascular disease.