Bayesian modeling application and optimization to demand forecasting

Abstract

Practices for optimal inventory management are a need at supply chains, especially for finished industrial products. A contribution to this logistic chain consists in finding efficient forecast of products demand, which permits to minimize cost inventory management, aspects that are more difficult in the presence of few historical data. This work proposal consists in the application of various Bayesian techniques with an optimization method, comparing its efficiency with MAPE indicator for demand forecasting, with few historical data. Results indicate that the expected value technique with an order 1 delay in the parameters, using Tabu metaheuristic shows the best accuracy in the forecast.