Research on Staff Scheduling in Sorting Centers Based on SARIMA-CEEMD-LSTM Model and Two-Stage Optimization Model

The present study developed an adaptive staff scheduling framework integrating demand time series prediction and operational optimization in sorting centers. A hybrid SARIMA (Seasonal Autoregressive Integrated Moving Average)-CEEMD (Complete Ensemble Empirical Mode Decomposition)-LSTM (Long Short-Term Memory) model was proposed, and a two-stage optimization model was established using genetic algorithms. Compared to standalone models, the hybrid prediction model has demonstrated a reduction in the Mean Absolute Error (MAE) by 12.8%. Concurrently, the optimization model provides managers with decision-making solutions at two levels of granularity. Empirical validation across 57 sorting centers confirms that this framework adapts to complex application scenarios, consistently performs prediction tasks, and flexibly adjusts optimization strategies.