基于深度信念网络的供应链柔性提升

Supply chain flexibility enhancement based on deep belief network

为了通过提高供应链柔性网络来提高企业运营效率,考虑供应端(节点中断)和需求端(需求量波动)的不确定性,运用含多隐层的机器学习模型——深度信念网络,构建了以中转点选择和流量分布优化为目标的供应链柔性网络模型,制定了网络训练步骤,并以某大型制造企业为例,在供应链柔性网络预处理的基础上,对比分析了BP神经网络与深度信念网络的流量预测精度。实例分析结果表明,深度信念网络克服了BP神经网络容易陷入局部最优的缺陷,比传统的BP神经网络预测精度更高、训练时间更短、学习能力更强,最大程度地缩短了供应链网络应对不确定风险的响应时间,提高了供应链网络柔性。

To improve supply chain flexibility network for raising the operational efficiency of enterprises,by considering the uncertainty of supply side（node interrupt）and demand side（demand fluctuation）simultaneously,and using the deep belief network that was machine learning model with multiple hidden layers,the model of supply chain flexible network was constructed.Different from the traditional BP neural network,the deep belief network overcame the defects of BP neural network in dealing with the nonlinear relation of complex function.Based on the actual data of a certain manufacturing industry,the contrast analysis showed that the deep belief network was more accurate than the traditional BP neural network.The training time of deep belief network was shorter,and the learning ability was stronger.It could shorten the response time of supply chain network to deal with uncertain risks,which improved the flexibility of the supply chain network.