Linfa project: significative steps

Our weekly Linfa project meeting today was quite an interesting one: we are running experiments to check whether a machine-learning based, prescriptive optimization model can beat stochastic optimization techniques. We think we found some evidence this might be the case. A simple k-nearest-neighbors based prescriptive model is delivering results which are very good based on two conflicting objectives: reducing stockouts (possibly to zero) and reducing stocks (to avoid waste).