AN OPTIMIZED ENSEMBLE MODEL FOR CARDIOVASCULAR WITH DIABETES DISEASE PREDICTION USING CGAN-AUGMENTED DATA

Cardiovascular disease (CVD) holds the position as the main killer worldwide in diabetic populations thus underlining the importance of accurate predictive tools. The inability of traditional statistical methods to adapt to data limitations alongside poor handling of clinical data imbalance leads to unsuccessful risk assessment. Deep learning solutions demonstrate promising results, yet they confront expensive computations and insufficient feature background understanding in addition to lacking interpretability features. The research introduces DFE-CVRP as a cardiovascular risk prediction system which merges expert models tailored for specific features and implements dynamic ensemble control with adaptive data balancing techniques. The performance evaluation determines if a lightweight ensemble model optimized dynamically improves CVD risk prediction results when processing structured clinical data. The method combines EfficientNet architectures which were optimized using Successive Halving and Population-Based Training methods and Conditional Generative Adversarial Networks to balance and improve feature diversity for the dataset. The performance of DFE-CVRP exceeds conventional machine learning techniques together with baseline deep learning architectures such as CCGLSTM when used on structured health databases. The algorithm reaches 98.2% accuracy and 97.8% precision combined with 98.4% recall while obtaining 98.1% F1-score and 98.6% AUC-ROC. The effectiveness of dynamic ensemble learning and data augmentation strategies for improved cardiovascular healthcare diagnosis has been confirmed through the study findings. The proposed predictive framework offers interpretability and scalability as well as affordable resource utilization that creates substantial value for future clinical decision systems leveraging patient-specific data.