SCDFUC · Supply Chain Demand Forecasting based on Unobserved Components models

The Supply Chain Management depends importantly on the predictions accuracy in the most of industries. These predictions are provided by the Forecasting Support System (FSS) in order to make decisions regarding departments like Marketing, Finance, Inventory, Distribution, Logistic, Human Resources and purchasing. In fact, these predictions are usually based on the mixture of forecasting statistical techniques, current economic situation, experience of the managers and the way that the FSS gathers these concepts. Nonetheless, there are current evidences that suggest a non efficient use of these systems and so, high costs are associated to these prediction errors. The present project will accomplish a thoroughly investigation about the possible sources of this inefficient use of the FSS by means of a collaboration with the Lancaster University Management School (LUMS). Thus, the different ingredients which act on the FSS will be analyzed in order to suit the main objectives of the organization in the best way. Firstly, we will analyze the different statistical methods which are candidates to do the forecasting task. We will focus on the Unobserved Components models developed in a State-Space framework, where novel hybrid techniques which use discrete and continuous time domains will be assessed in combination with efficient recursive estimation techniques like Kalman Filter and Fixed Interval Smoothing. Secondly, a study about the influence of the current economic situation on our forecasts will be accomplished. This investigation will be carried out from a new point of view about the business cycle, where adaptive nonlinear techniques which come from the control literature will be used to allow us look into the time-varying behaviour of the business cycle frequency. Finally, all the aforementioned points will be gathered with the manager’s judgement in an ideal Forecasting Support System.