Demand Forecasting of Individual Probability Density Functions with Machine Learning

Demand forecasting is a central component of the replenishment process for retailers, as it provides crucial input subsequent decision making like ordering processes. In contrast to point estimates, such conditional mean underlying probability distribution, or confidence intervals, complete density functions allows investigate impact on operational metrics, which are important define business strategy, over full range expected demand. Whereas metrics evaluating estimates widely used, methods assessing accuracy predicted distributions rare, and this work proposes new techniques both qualitative quantitative evaluation methods. Using supervised machine learning method "Cyclic Boosting", individual can be that each prediction fully explainable. This particular importance practitioners, avoid "black-box" models understand contributing factors prediction. Another aspect in terms explainability generalizability demand limitation influence temporal confounding, prevalent most state art approaches.