At present, the most prevalent disease causing deaths worldwide is Heart Disease (HD). The accurate diagnosis of heart disease is important in medical science, where Machine Learning models can play a key role in enhancing predictive accuracy. As modern machine learning techniques depend on large datasets to obtain stable results, generation of synthetic data has become important in the field of medical domain to overcome this lack of data. This study investigates the use of Gen AI based synthetic data generation technique overcomes the challenges posed by the limited size of the dataset. We employ the CTGAN (Conditional Tabular GAN) model to generate synthetic data as it is designed for tabular data handling both continuous and categorical features effectively. Additionally, we employed Genetic Algorithm (GA) and Particle Swarm Optimization (PSO) to select the most relevant optimal features, as both are computationally efficient and require relatively fewer parameters to adjust. These features were then used to train multiple classifiers, including K-Nearest Neighbors (KNN), Random Forest, Decision Tree, Support Vector Machine (SVM), and a Deep Learning (DL) model with two hidden layers. The proposed approach’s efficiency is validated using multiple evaluation metrics on two datasets such as Cleveland and Statlog. The results suggest that the DL model achieved the highest accuracy, outperforming other Machine Learning (ML) models. Thus, integrating synthetic data and feature selection with deep learning effectively improves HD prediction.