Artificial Intelligence in Logistic Warehousing: A Case Study on Stock Management Optimization

Abstract

The rapid advancement of Industry 4.0 technologies has transformed warehouse operations, with artificial intelligence emerging as a pivotal tool for optimizing inventory management. This paper proposes an artificial neural network based forecasting model designed to address two critical challenges in warehouse logistics: overstocking and stockouts. Using a case study of a Moroccan beverage distributor, the model was trained on two years of operational and contextual data, incorporating advanced preprocessing, feature engineering, and performance evaluation using MSE, MAE, and R² metrics. Comparative experiments demonstrate that the proposed ANN outperforms ARIMA, Long Short-Term Memory networks, and XGBoost, achieving an R² of 0.89 and reducing MAE by up to 28% compared to traditional methods, while maintaining low computational requirements. Enhanced by the integration of external variables—such as promotional campaigns, weather conditions, and economic indicators—the model achieved an improved R² of 0.93 in preliminary tests. Results also reveal quantifiable operational benefits, including a 22% reduction in order preparation time, an 18% decrease in labor hours, and a 15% reduction in storage errors. These findings highlight ANN’s capacity to capture complex, non-linear demand patterns and its suitability for scalable, real-time industrial deployment, positioning it as a strategic enabler for resilient, data-driven warehouse management in the Industry 4.0 era.