Designing the smart warehouse: key automation criteria for sustainable and scalable operations

Purpose- Warehouses face challenges due to rising demands for efficient, sustainable, large-scale operations, making automation essential for enhancing processes and reducing costs. This paper compares three automation models LSTM, Prophet, and Logistic Regression that can improve warehouse management, particularly in sales forecasting and reorder prediction. Methodology- Using actual warehouse sales data from Kaggle, time-series models (LSTM and Prophet) were built for daily sales forecasting and Logistic Regression for reorder quantities on each item. The models evaluated each model's ability to project warehouse sales and determine replenishment timing. Findings- Results suggest that LSTM provided better forecasting results than Prophet, with lower MSE, RMSE, and MAE values, modelling both short-term volatility and long-term trends. Logistic Regression showed high accuracy and decent precision, though low recall suggests it missed many reorder cases. Conclusion- While LSTM models can improve decision-making in warehouse management, further development of classification models is essential to enhance reorder prediction accuracy, increase recall, and prevent stockouts.