Heart diseases are still among the leading causes of death in the global population, which confirms the relevance of timely and correct diagnostics. This research paper explores the development of a machine learning model for predicting heart disease using Recurrent Neural Networks (RNNs) and Long Short-Term Memory (LSTM) units. The research uses the UCI Heart Disease dataset for analysis with the following data pre-processing – normalization and features selection with the help of k-Nearest Neighbors (k-NN) algorithm. In line with this, the proposed RNN-LSTM model reached the highest accuracy of 98%, precision 97%, recall 100%, F1-measure 98%, and Receiver Operating Characteristic-Area Under the Curve (ROC-AUC) measure of 98% These percentages show how effective and reliable the model in predicting heart disease to be considered in the clinical decision support system. The conclusions of this project indicate that better and timely identification of the issues would mean less load on the healthcare systems and better results for patients. Possible future work involves incorporating more features into the model and adopting a mixed method approach to improve the level of prediction and transferability. This study is a testament to how machine learning is revolutionizing healthcare, opening up possibilities for even further development of diagnostic capabilities.