Multi-layer perceptrons for on-line lot sizing

Considers an on-line lot sizing problem with overtime. The authors develop a two-stage decision procedure for this problem. In the first stage an MLP classifies the decision situation. It is in this stage that uncertainties are taken into account. The outcome of the first stage is used as input for the second stage, in which a detailed production plan is calculated. The proposed approach combines the classification and pattern recognition abilities of MLPs with traditional deterministic analysis. The authors give a brief introduction in MLPs and supervised learning and the on-line lot sizing problem is formulated. Based on results for the deterministic finite horizon problem the authors derive a two-stage strategy for the on-line lot sizing problem. Finally in they discuss some results.