Forecasting Drug Dispensation Using Machine Learning: A Case Study on a Large Public Hospital

Accurate medication demand forecasting is essential for optimizing inventory management in large government hospital dispensations. This study evaluates four machine learning models: Support Vector Regression (SVR), Extreme Gradient Boosting (XGBoost), Long Short-Term Memory (LSTM), and Gated Recurrent Units (GRU), comparing their forecasting performance against the hospital’s current method, Simple Moving Average (SMA). The evaluation focuses on the top 20 drugs with the highest total amounts received from the drug warehouse for each pharmaceutical room, ensuring that critical medications are prioritized. Using Mean Absolute Percentage Error (MAPE) as the primary evaluation metric and Symmetric Mean Absolute Percentage Error (SMAPE) as a secondary metric, results indicate that the SVR model could enhance inventory management efficiency and reduce medication shortages or overstock scenarios.