Optimizing Pharmaceutical Supply Chains using a Hybrid CNN-LSTM Model for Accurate Medicine Demand Forecasting
P. Priyadharshini, S. Keerthana, G. Keerthana Devi · 2025
In recent years, predicting medicine demand in real-time has become a critical challenge for hospitals and pharmacies, especially with unpredictable surges due to public health emergencies, seasonal diseases, and supply chain disruptions. Traditional forecasting models struggle to manage non-linear and dynamic patterns in pharmaceutical consumption, often leading to overstocking or critical shortages. These challenges are exacerbated by the lack of integration between temporal data trends and real-world influencing factors such as public holidays, climate, and population density. To address these limitations, this work proposes an intelligent medicine demand prediction system using a hybrid CNN-LSTM deep learning architecture. CNN layers are employed to extract robust local temporal features from historical sales and inventory datasets, while LSTM layers capture long-term dependencies and sequence trends. This combination enhances the model’s ability to accurately predict fluctuating demand across various time scales. The system is trained and validated on real-world datasets encompassing multiple medicines, seasonal patterns, and geographic variations. Key features of this methodology include advanced data preprocessing with handling of missing values, feature scaling, and categorical encoding; classification of demand levels into categories for effective inventory decisions; and a robust evaluation using metrics like accuracy, precision, recall, and F1-score. The model is further benchmarked against traditional forecasting techniques to demonstrate superior performance. The proposed solution enables proactive inventory planning, minimizes wastage, and ensures continuous availability of essential medicines in healthcare systems.