This study proposes a new method to improve the efficiency of drug discovery by repurposing existing drugs, aiming to reduce the time and costs associated with traditional drug development processes, which can span 10 to 15 years and cost billions of dollars. Current approaches focus on leveraging heterogeneous data, such as drug-protein and disease-protein interactions, to construct complex networks that link drugs, proteins, and diseases. However, a significant challenge is the imbalance in data, where numerous unconfirmed potential drug-disease interactions (the majority class) outnumber approved drugs (the minority class), severely impacting the predictive performance of machine learning models. Previous attempts to address this issue have shown limited success. This study introduces a novel approach that integrates meta-paths in heterogeneous information networks with data balancing techniques to tackle this imbalance. Experimental results demonstrate that the proposed method enhances model performance and reliability in identifying new relationships between drugs and diseases. This research represents a promising advancement by leveraging network-based strategies and data balancing techniques to facilitate the rediscovery of drug applications, thereby potentially revolutionizing the pharmaceutical industry’s approach to drug development.