SARIMAX-FTS Hybrid Model for Food Input Demand Forecasting

Forecasting demand in environments subject to seasonality, operational variability and the influence of external factors represents a relevant challenge for inventory management. This study proposes a hybrid SARIMAX-FTS model applied to the forecast of demand for food inputs for the bakery segment. The model combines the statistical capacity of SARIMAX in modeling trend, seasonality, and exogenous variables with the fuzzy representation of uncertainty through the Fuzzy Time Series (FTS). Past demand, inventory, price and seasonal events were used as variables. The models were evaluated using MAE, MMR, MAPE and R². The combination between the classical statistical model SARIMAX and the model with FTS computational intelligence allowed to expand the interpretability of the forecast by explicitly incorporating seasonal effects and exogenous variables and, simultaneously, capturing nonlinear relationships and uncertainties inherent to the series through fuzzy modeling.