O uso de modelos de previsão de demanda em uma indústria do ramo alimentício do agreste alagoano

Abstract

The objective of this study was to analyze the performance of quantitative demand forecasting methods based on time series in a food industry located in the Agreste region of Alagoas. This development is justified by the growing need of industries for tools that allow them to expand their presence in the market. The choice of a method with lower margin of error is of utmost relevance for the strategic planning of organizations. The model is based on time series analysis, examines the stationarity of the dataset through the coefficient of variation, and observes the components of the series: trend, seasonality, and error, finally generating the forecast. The research method applied was simulation, using forecasting models such as the Exponential Smoothing with Error, Trend, and Seasonality (ETS) method and the Box-Jenkins Autoregressive Integrated Moving Average (AUTO.ARIMA) method. Due to the high capacity of the models in simulating various scenarios, the choice of the method that best fit the dataset of this research was made. The results obtained demonstrated that the model with the highest degree of accuracy was ETS for beans, with better results in error coefficients: MAPE (12.338), MAE (51.893), and MASE (0.77440), and AUTO.ARIMA for rice, with coefficients: MAPE (22.527), MAE (31.942), EM (-0.574), RMSE (42.635), and MASE (0.734). The purpose of these models is to bring greater accuracy to organizational data, as well as to generate reliable and realistic information to foster the company's development through the use of modern and user-friendly forecasting models.