In the context of modern medicine, where polypharmacy is prevalent, particularly among older adults, the potential for hazardous drug-drug interactions (DDIs) is a significant concern. These interactions can lead to severe adverse reactions, with consequences ranging from debilitating health complications to fatality. Previously, distinct machine learning and deep learning algorithms were used to predict DDIs using gold-standard datasets. These methods demonstrate remarkable accuracy but lack precision due to a high false positive rate. Therefore, there is a crucial need to enhance the models’ ability to reduce the false positive rate. This paper introduces a deep learning-based framework named InteractNet to enhance precision in DDI predictions. This framework employs 1D convolutional neural networks with streamlined parameterization in the convolutional layer for multi-scale feature extraction, bidirectional long short-term memory to capture drug feature interactions, and an attention mechanism to concentrate on key data features for precise predictions. The efficacy of InteractNet is thoroughly examined through five performance metrics, namely accuracy, precision, true positive rate, harmonic mean, and ROC-AUC score. Results indicate that InteractNet achieves the highest testing accuracy of 94.29 ± .0088, a precision of 94.23 ± .0023, a true positive rate of 94.49 ± .0037, a harmonic mean of 94.21 ± .0029, and a ROC-AUC score of 96.24 ± .0032. These outcomes surpass those of existing state-of-the-art methods, underscoring the profound potential of deep learning to enhance medical safety.