Machine learning-based forecast of procurement budgets for low-value consumables

Procurement budget forecasting for low-value consumables is critical for corporate cost control.Addressing the limitations of traditional statistical methods in handling demand fluctuations, this study proposes a hybrid machine learning model integrating seasonal decomposition with random forest.Validated using public supply chain datasets, this model reduces the mean absolute percentage error of budget forecasts to 12.3%; significantly outperforming the autoregressive integrated moving average model (18.5%) and linear regression methods (16.2%).Experimental results demonstrate that by integrating temporal characteristics of historical procurement data with external influencing factors, the model achieves a coefficient of determination of 0.89 on the test set.The weighted mean absolute percentage error metric is reduced by approximately 35% compared to baseline methods, providing enterprises with a more precise budget forecasting tool for procurement decision-making.