Previsão de demanda para materiais de construção a seco: uma análise por séries temporais

Abstract

The study addresses demand forecasting for dry construction materials, focusing on the Light Steel Framing system using time series analysis. The dry construction sector, which is expanding in Brazil, faces challenges in supply chain management due to demand variability and the specific characteristics of the materials used. This work is justified by the importance of demand forecasting for production efficiency and the sustainability of this market. Several time series models were applied to identify the most accurate one in terms of the lowest set of errors. Simpler models showed high errors, while more complex models, such as SARIMA, did not outperform intermediate alternatives due to the short historical series and low seasonality of the data. The additive Holt-Winters model achieved the best performance, with a MAPE of 5.72%, MAD of 5,203.40, and RMSE of 6,434.92, proving to be the most suitable for representing demand patterns.