Effective Procurement and Demand Forecasting for Ready-Mix Concrete Plants

Abstract

Accurate forecasting of Ready-Mix Concrete (RMC) demand helps optimize procurement, reduce waste, and lower environmental impact. This study presents an AI-based system that predicts daily demand using 18 operational features and recent 5,000 data points (80:20 split). Multiple models were tested, where LSTM performed poorly (MAPE 145.10%), while a hybrid model combining XGBoost (0.35) and Random Forest (0.c5) achieved the best results (RMSE 3.c8 m³, MAPE 1.55%). Feature analysis shows cement density (32.81%) as the most influential factor. The system also estimates CO₂ emissions, enabling cost-efficient and sustainable procurement decisions, demonstrating the effectiveness of hybrid ensemble learning in construction forecasting.