<p>One of the leading causes of morbidity and mortality among non-communicable disease worldwide is heart disease. Recently, the field of intelligent healthcare has developed numerous deep learning and machine learning frameworks for predicting cardiovascular disease. Nevertheless, most current methods have failed to achieve improved prediction accuracy for heart disease due to a lack of suitable, data-driven prediction methodologies. In this research, a new deep learning-based framework is proposed for heart disease prediction, motivated by existing problems. The primary goal of this research is to develop a state-of-the-art CNN model with an advanced architecture for effective decision-making and precise disease prediction. Initially, the suggested method collects data on heart disease from various publicly accessible data sources. The acquired data is then preprocessed using three different techniques. The dataset’s significant features are extracted using MobileNetV2. The best features are selected using the Improved Particle Swarm Optimization (IPSO) technique. Finally, the Enhanced AlexNet (EAlexNet) model was used to predict cardiovascular disease. The UCI Heart Disease and Framingham benchmark datasets are used to validate the proposed approach, yielding accuracy rates of 98.96% and 98.74%, respectively. Finally, a comparison study demonstrates that the proposed predictions are more accurate (with a smaller feature set) than those of existing state-of-the-art methods. Therefore, the proposed approach for early heart disease diagnosis is highly reliable and feasible in a real-world setting. The suggested model can assist radiologists and doctors in more reliably diagnosing cardiac patients by increasing the system’s effectiveness and returning the most practical solution among all input prediction models while considering performance criteria. The proposed deep learning method is emerging, with the potential to predict early heart disease and lower heart disease mortality.</p>