ON THE LIAISON BETWEEN FORECASTING AND PERIODIC INVENTORY CONTROL: APPLICATION OF SARIMA MODELS TO INTERMITTENT DEMAND PROFILES

The liaison between forecasting and inventory control represents a promising research field even if they have been often considered as independent problems. Otherwise, this paper follows a recent set of works that proved the importance of simultaneously considering both aspects. As a consequence, the choice of the best-performing approach should not consider only the accuracy of forecasts, but also the effects of forecasting methods on re-order policies. In particular, this paper deals with different periodic inventory control methodologies, with different lead times and safety stocks. The application to intermittent demand patterns is studied. The best Seasonal Autoregressive Integrated Moving Average (SARIMA) forecasting models are initially automatically identified by the TRAMO-SEATS software and results obtained are subsequently used for tuning periodic review inventory control approaches. The experimental analysis is performed on a real data set of 40 time series. A discrete-event simulation finally shows the obtained effects. Two features of this paper are remarkable. The former is that the intermittent data set is not pseudo-random generated and thus assumptions on the best-fitting demand distribution are not required. The latter regards the full automation of the TRAMO-SEATS software, which indicates its ability of being applied also in real industrial environments.