Salvato in:
| Autori principali: | , |
|---|---|
| Natura: | Recurso digital |
| Lingua: | |
| Pubblicazione: |
Zenodo
2025
|
| Accesso online: | https://doi.org/10.5281/zenodo.17812792 |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
Sommario:
- <p>Supply chain management faces increasing com-<br>plexity due to demand uncertainty, lead time variability, and<br>operational costs. This paper presents a deep reinforcement learn-<br>ing (RL) framework using Q-learning for optimizing inventory<br>management decisions in supply chain systems. We develop an<br>intelligent agent that learns optimal ordering policies by bal-<br>ancing holding costs, stockout penalties, and ordering expenses.<br>Using real-world retail transaction data, our experimental results<br>demonstrate that the RL-based approach achieves 23.4% cost<br>reduction compared to traditional constant ordering policies<br>and 15.8% improvement over threshold-based heuristics. The<br>proposed system achieves a 94.2% service level while maintaining<br>average inventory levels 18% lower than baseline methods.<br>Our findings suggest that RL-based approaches offer significant<br>potential for autonomous supply chain optimization, particularly<br>in dynamic demand environments.</p>