CNN-LSTM Plus MILP Framework for Coffee Supply Chain Optimization
June 9, 2026
A hybrid CNN-LSTM demand forecasting model achieves MAE of 22.87 and R² of 0.90 on a coffee sales dataset, outperforming deep learning baselines by 12% and classical methods by over 30%. Forecasts feed a tri-objective MILP minimizing cost, carbon emissions, and freshness simultaneously in a closed-loop supply chain.
HOW THIS AFFECTS YOU
●
researcherThe two-phase forecast-then-optimize architecture is a reusable pattern for agri-food supply chain problems, with the MILP formulation handling competing sustainability objectives jointly.