DEA de GSI
Transcription
DEA de GSI
Non-stationary demand forecasting Année et Spécialité du Master Recherche : M2- OSIL Titre du sujet proposé : Non-stationary demand forecasting by non-overlapping temporal aggregation Nom de l’organisme ou de l’entreprise qui propose le sujet : Nom du resp. du sujet (encadrant) : Bahman ROSTAMI-TABAR Numéro de téléphone du resp. du sujet : 01 41 13 16 59 Email du resp. du sujet : [email protected] Contexte et présentation du sujet : Aggregation across time, also called temporal aggregation, refers to the process by which a low frequency time series (e.g. quarterly) is derived from a high frequency time series (e.g. monthly) (Nikolopoulos et al. 2011). As shown in Figure 1, this is achieved through the summation (bucketing) of every m periods of the high frequency data, where m is the aggregation level. In nonoverlapping temporal aggregation, (Figure 1) the time series are divided into consecutive nonoverlapping buckets of time where the length of the time bucket equals the aggregation level. The aggregate demand is created by summing up the values inside each bucket. The number of aggregate periods is [N/m], where N is the number of the original periods, m the aggregation level and the [x] operator returns the integer part of x. As a consequence the number of periods in the aggregate demand is less than the original demands. Figure 1: Non-overlapping temporal aggregation (from weekly to monthly data) In our previous work, we evaluate the impact of temporal aggregation on stationary demand processes(Rostami-Tabar et al. 2013, Rostami-Tabar et al. 2014). Compared to stationary demand processes, nonstationary processes have not been considered in this context, although there is evidence that most of the forecasting and inventory control problems occur in situations where demand is nonstationary(Sbrana and Silvestrini 2013). Objectifs et résultats attendus de l’étude: The objective of this study is to evaluate the impact of non-overlapping temporal aggregation on demand forecasting. We start by conducting the mathematical analysis to reveal the benefits of temporal aggregation for nonstationary process compared with stationary one. Next, we do a simulation analysis to ensure the results of analytical evaluation. Finally, a real data sets ca, be used to assess the validity of results on real life exempts. Connaissances requises éventuelles : Methodologies : Mathematical modeling, Simulation Tools : Matlab or R Perspectives éventuelles (publications? stage de mémoire de recherche ? thèse ? embauche ?) : There will be a potential publication intended for: International Journal of Production Economics (IJPE)/ IMA Journal of Applied Mathematics (IMA) References: Nikolopoulos, K., Syntetos, A.A., Boylan, J., Petropoulos, F. & Assimakopoulos, V., 2011. An aggregatedisaggregate intermittent demand approach (adida) to forecasting : An empirical proposition and analysis. Journal of the Operational Research Society, 62 (3), 544-554. Rostami-Tabar, B., Babai, M.Z., Syntetos, A. & Ducq, Y., 2013. Demand forecasting by temporal aggregation. Naval Research Logistics (NRL), 60 (6), 479-498 Available from: http://dx.doi.org/10.1002/nav.21546. Rostami-Tabar, B., Babai, M.Z., Syntetos, A. & Ducq, Y., 2014. A note on the forecast performance of temporal aggregation. Naval Research Logistics (NRL). Sbrana, G. & Silvestrini, A., 2013. Forecasting aggregate demand: Analytical comparison of top-down and bottom-up approaches in a multivariate exponential smoothing framework. International Journal of Production Economics, 146 (1), 185-198 Available from: http://www.sciencedirect.com/science/article/pii/S0925527313002922.