Combination forecasting method using Bayesian models and a metaheuristic, case study

Planning of demand forecasting for perishable products is important for any type of industry that manufactures or distributes, especially if it has a seasonal behavior and a difficult to predict variability. This paper proposes a metaheuristic based on Ant Colony Optimization (ACO) for the combination of forecasts of multiple products, based on three models: Mixed Linear Model (MLM), Bayesian Regression Model with Innovation (BRM) and Dynamic Linear Bayesian Model (BDLM), which are part of the proposed combination whose process is based on minimizing the Mean of Absolute percentage Error (SMAPE) indicator. It is found that the BDLM and BRM methodologies obtain good results on an individual basis, being better BRM, however, the ACO algorithm designed yields a better result, facilitating an adequate prediction of the demand of several products of a company in the meat buffer sector. Keywords: statistics and probability; forecasts; optimization theory; Bayesian statistics.

• Autores: Higuita Alzate, David; Valencia Cárdenas, Marisol; Correa Morales, Juan Carlos
• Tipo de recurso: Article of investigation
• Fecha de publicación: 2018
• Institución: Tecnológico de Antioquia
• Repositorio: Repositorio Tdea
• Idioma: spa
• OAI Identifier: oai:dspace.tdea.edu.co:tdea/2843
• Acceso en línea: https://dspace.tdea.edu.co/handle/tdea/2843
• Palabra clave: Forecasts; Productos perecederos; Perishable products; Produto perecível; Produit périssable; Statistics and probability; Estadística y probabilidad; Pronósticos; Optimization theory; Teoría de optimización; Bayesian statistics; Estadística Bayesiana
• Rights: openAccess
• License: https://creativecommons.org/licenses/by-nc-nd/4.0/