Can learning intelligence outperforms information sharing in supply chain performances - an order arrival prediction perspective

To improve competitive advantage and operational performances of supply chains, we implement multi-agent supply chain modeling with learning capability to predict order arrival times that manufacturers can pre-produce to shorten order lead time for downstream customer orders. As order lead time is reduced, bullwhip effect of supply chains would also be minimized. Two kinds of learning agents are embedded in traditional supply chains to learn from past experiences to predict next order arrival time. We use back propagation neural networks and an order arrival pattern matching (OAPM) algorithm with belief set models for the prediction. The performances are compared with the traditional supply chain and the VMI-based supply chain. Results show that even with tailored learning intelligence, the VMI-based supply chain still performs better than the others. However, the two supply chains with learning agents outperform the traditional supply chain. This implies that learning intelligence can assist in predicting order arrival times, but information sharing seems to do it even better.