Forecasting Models Selection Mechanism for Supply Chain Demand Estimation

Abstract

The aim of this work is to present a selection mechanism of forecast models to contribute to demand estimation in a supply chain. At present, to estimate a product future demand, several forecast models based on historical information - quantitative and qualitative- are used. When companies face this situation, they select a group of forecast models (usually based on a visual basis of the time series), then estimate, and with the forecast error measurement criteria decide which the best method is. But they always have to estimate over all the selected forecast models. Based on that, this paper introduces an alternative methodology to estimate the best-forecast model without the need to estimate all the forecast models or complement with another technique (visual). To do so, the main theoretical fundaments associated to this new methodology are addressed, and then the methodology itself is presented in order to be applied in two real cases of Chilean companies to finally conclude the results of the described mechanism.