Enhancing Stockpile Management Through Deep Learning with a Focus on Demand Forecasting and Inventory Optimization

Stockpile management is essential for industries to efficiently meet customer demands while minimizing costs. Traditional methods like the Economic Order Quantity (EOQ) model and ARIMA often struggle with adaptability and accuracy, leading to inefficiencies such as overstocking or stockouts. This study proposes using Long Short-Term Memory (LSTM) networks to improve demand forecasting in stockpile management. LSTMs are designed to handle sequential data, capturing complex patterns and long-term dependencies that conventional methods miss. The LSTM-based approach achieved up to a $\mathbf{0. 3 0 \%}$ improvement in forecasting accuracy, resulting in more precise inventory control and better alignment with actual demand. The proposed system, termed Deep Inventory Management (DIM), incorporates LSTM models that adapt to changes in demand by considering factors such as seasonality and trends. This adaptive approach allows for proactive inventory predictions, enhancing decision-making in procurement and stock levels. Key abbreviations include LSTM (Long Short-Term Memory), EOQ (Economic Order Quantity), and DIM (Deep Inventory Management). By aligning inventory more closely with demand, this deep learning approach enhances operational efficiency, reduces costs, and improves customer satisfaction, making it a valuable advancement over traditional inventory management methods.