HYBRID FEATURE OPTIMIZATION AND COMPUTING PLATFORMS FOR CARDIAC RISK PREDICTION

Abstract

Cardiovascular disease remained has one among the major causes for the fatality globally, accentuating the significance of premature prediction and risk detection. Latest break-throughs in Machine learning techniques have been increasingly deployed to medical dataset to assist medical practitioners in decision making. Several research have developed predictive architecture utilizing classification algorithms along with that feature selection techniques, hybrid learning models and comparative analysis of multiple algorithms are also incorporated to find out bottlenecks and improve prediction capability. Regardless of these advancements, prior researches depend on relatively small or spatial-specific datasets, lacking effective feature optimization, increased computational complexity, reduced interpretability for clinicians. These difficulties emphasize need for proficient integrated framework which productively deduct feature redundancy, enhance model generalization and prediction accuracy. The proposed framework aims to develop an interpretable and accurate cardiovascular disease prediction system by combining feature optimization techniques with machine learning classification model. The developed architecture focuses on recognizing most prominent clinical attributes, decrease redundant and correlated variables and used optimized features to train robust predictive model for early detection of CVD. Initially, data preprocessing and normalization are performed on dataset. Dimensionality is decreased by applying Principal Component Analysis (PCA), succeeded by Improved Dragonfly Optimization algorithm (IDFOA) to choose most related features. The optimized features are later utilized to train deep learning algorithm like predictive model and gradient boosting machine for disease prediction. The experimental analysis illustrates that integrated optimization approach significantly enhances CVD disease prediction performance relative to baseline models. The combination of PCA and IDFOA valuably decreases redundant features while enhancing sensitivity, and Gini coefficient. (The Randon Forest) accomplishes  best predictive performance. Overall, the designed framework supplies proficient and reliable method for CVD prediction and supports intelligent clinical decision support systems for early CVD risk assessment.