Heart Failure affects millions worldwide, and early identification is crucial to prevent serious complications. Current diagnostic approaches depend mostly on human interpretation and are frequently inaccurate. Artificial Intelligence models are efficient for estimating heart failure progression since they provide healthcare professionals with more accurate and automated tools. This study aims to train and evaluate DL models (1D-CNN, DNN, ANN) on raw and feature-engineered cardiac arrest detection datasets. The performance of this model was first evaluated using heart disease data. On the raw dataset, ANN outperformed with the highest Accuracy of 80% and AUC of 82.74%. After feature engineering using XGBoost, ANN’s performance increased with an Accuracy of 95.08% and an AUC of 93.99%, indicating that feature engineering improves prediction. These findings emphasize the potential based on AI models in developing more reliable and automated diagnosis techniques for cardiac diseases. Future research will integrate real-time patient monitoring data with improved predictive models to detect developing health issues and allow for more accurate early diagnosis. This technique can transform heart failure therapy by providing personalized and more effective medicines, improving clinical decision-making and patient outcomes.