This study aims to improve the accuracy of clinical drug interaction warnings in CDSS by using deep learning techniques. The majority of traditional CDSS systems have rule-based algorithms that generate numerous false positives, which leads to desensitization of the clinician other words, “alert fatigue”-and can be derogative from patient safety. With this in mind, this paper puts forward a deep learning model that provides richer, more accurate drug interaction alerts through the use of substantial amounts of EHR and pharmacological databases, including patient-specific variables such as demographics, medication history, and comorbidities. The methodology of the model involved developing a multilayer neural network trained with a wide dataset. Some of the preprocessing steps of data involved cleaning and normalization of the datasets, and feature selection to ensure consistency and accuracy in the system. The model’s performance was judged on its accuracy, precision, recall, and F1 score. The proposed system was then compared with a more traditional rule-based system. This research aims to minimize false-positive alerts and enhance the overall reliability of CDSS while predicting drug interactions. The results are very promising and can be seen with a highly improved accuracy and considerably reduced false positives; deep learning contributes to the potential of enhancing medication safety and supporting clinicians in decision-making in real-time clinical environments.