A REVIEW OF ARTIFICIAL INTELLIGENCE MODELS FOR
DEMAND FORECASTING IN HEALTHCARE SUPPLY CHAIN

Sourish, Shreevathsa, Mihir R Shandilya, Ayisha Afra, Kathija Shafa

Healthcare supply chains, especially those handling essential medicines - still struggle with a basic but critical problem: making sure the right medicines are accessible when they’re needed. In reality, this comes down to two issues: unreliable demand forecasts and inefficient inventory practices. Traditional statistical methods are widely used, but they tend to fall short when demand shifts due to seasonality, disease outbreaks, or policy changes. More advanced machine learning and deep learning models help capture these patterns better, but they’re often not well incorporated with real-world decision-making.
This review looks at a variety of forecasting approaches, including statistical models like SARIMA, machine learning methods such as Random Forest and XGBoost, and deep learning techniques like LSTM. Each approach is examined based on what it does best- whether it’s identifying trends, handling nonlinear relationships, or learning long-term demand patterns. Their performance is also discussed using standard evaluation parameters such as MAE, RMSE, and MAPE, which help compare their practical effectiveness.
Beyond forecasting, this paper also speaks about how these predictions can be utilised to guide inventory decisions. This includes basic but essential parameters like safety stock levels and reorder points. Studies that attempt to link forecasting with inventory management are reviewed critically to understand their usefulness in real-world settings. Overall, the review highlights a clear gap: while forecasting models are becoming increasingly advanced, fully integrated solutions that augment prediction with action are still limited. Addressing this gap is essential. Future work should concentrate on building practical, data-driven systems that not only enhance forecasting accuracy but also support better inventory decisions. This can help decrease waste, prevent stockouts, and ultimately improve access to essential medicines while enhancing the resilience of healthcare supply chains.

Keywords –Essential medicines; Demand forecasting; Healthcare supply chain; Machine learning; Deep learning.