Advancing Demand Forecasting for Material Requirements Planning in Food Manufacturing: A Hierarchical Forecasting Approach

This study investigates the enhancement of demand forecasting accuracy within Material Requirements Planning (MRP) in the food manufacturing industry by employing hierarchical forecasting. The research evaluates various forecasting models and reconciliation approaches by employing hierarchical forecasting which segments time series data into multiple aggregation levels. The models tested include statistical techniques; autoregressive integrated moving average (ARIMA), exponential smoothing (ETS), prophet and machine learning techniques; linear regression (LR), k-nearest neighbors (KNN), support vector regression (SVR), random forest (RF), and extreme gradient boosting (XGBoost). Among the reconciliation approaches; bottom-up, ordinary least squares (OLS), and minimum trace (MinT), Random Forest combined with OLS showed the best performance. A unified forecasting model for total product demand improved accuracy by 20-25% compared to aggregating individual forecasts. These findings underscore the value of advanced forecasting methods in optimizing production planning offering significant practical and academic contributions to the food manufacturing sector.