Machine learning (ML) has revolutionized healthcare, offering precise tools for disease prediction, notably in heart conditions (Jamthikar et al. in Cardiovasc Diag Therapy 10(4):919, 2020, [7]; Siontis et al. in Nat Rev Cardiol 18(7):465–478, 2021, [17]). Timely detection is crucial for effective treatment, with ML playing a key role. ML models analyze extensive datasets, including patient history and imaging data, to predict heart disease based on factors like age, gender, and cholesterol levels. For instance, electrocardiogram (ECG) analysis detects heart rhythm abnormalities. ML complements traditional methods, enhancing accuracy and efficiency without replacing them. Ejection fraction (EF), crucial for assessing heart function and guiding treatment, is predictively modeled using ML techniques. Our approach optimizes model size and accuracy by representing entire videos as single-image-sized data through auto-encoders. Employing deep neural networks and ensemble methods, our model achieved high accuracy and low error rates, improving EF prediction for early disease detection and treatment optimization. Regular EF monitoring assesses treatment efficacy. Our research has reduced model complexity, yielding a 20-million-parameter model with a 4.2 Mean Absolute Error (MAE) in EF prediction. ML not only enhances accuracy and efficiency but also provides data-driven insights, accelerating diagnostics and improving treatment outcomes.