AI Demand forecasting for inventory management

Inventory management literature and software often assume that demand patterns remain constant over time, which is unrealistic in practice due to seasonal fluctuations, among other factors. Traditional forecasting models often do not take uncertainties into account, even though they are crucial for efficient inventory decisions. This project focuses on a predict-then-optimize approach, better integrating demand forecasting into inventory management to minimize costs due to shortages or overstock. In addition, we investigate whether a more integrated method, where decisions are derived directly from data, yields better results than the traditional approach.