The Application of Federated Active Learning in Supply Chain Demand and Supply Forecasting
نویسندگان
1 PhD, Full Professor, Department of Management, Te.C., Islamic Azad University, Tehran, Iran.
2 PhD, Assistant professor, Department of Mathematics and Statistics, Qe.C., Islamic Azad University, Qeshm, Iran.
3 PhD Student, Department of Industrial Management, Qe.C., Islamic Azad University, Qeshm, Iran.
4 PhD, Assistant professor, Department of Computer Engineering, Qe.C., Islamic Azad University, Qeshm, Iran.
doi
10.22034/kps.2025.533701.1236چکیده
This study explores the application of federated learning in demand forecasting for decentralized supply chains, focusing on enhancing data privacy, forecasting accuracy, and computational efficiency. Federated learning allows multiple nodes, such as retailers and warehouses, to collaboratively train machine learning models without sharing sensitive data. The integration of active learning further improves the model’s accuracy by prioritizing the most informative data points, thereby reducing training time and communication costs.The results demonstrate that the federated learning model significantly outperforms traditional centralized models in terms of accuracy, with a 45% improvement in Mean Absolute Error (MAE) and a 31% improvement in Root Mean Square Error (RMSE). Moreover, the federated model reduces computational overhead by 35% and enhances privacy, achieving lower epsilon (ε) values, indicating stronger privacy guarantees. These findings suggest that federated learning is a viable and effective solution for real-time demand forecasting in complex and decentralized supply chains.Future work can build on this approach by integrating additional privacy-preserving techniques and expanding its application to other areas of supply chain management. The study contributes to the growing body of knowledge on the use of artificial intelligence in supply chain optimization, offering a scalable and privacy-preserving alternative to traditional forecasting methods.
