链路预测下能源供应链网络合作演化机制研究

摘要

Using attribute information of a given network structure or nodes of a supply chain to predict the possibility of cooperation with an unlinked enterprise is key to link prediction being applied to supply chain networks.As such, we predicted side-connected evolutions of network cooperation in energy supply chains by utilizing a theoretical framework and evaluation method for link prediction and five similarity indexes.Our results show that when the attribute of the network structure of a supply chain is used as a single similarity index, the predictive effect of the RWR index is the best.Further, the prediction accuracy of the coupling index is remarkably higher than the single index.Here, the coupling effect between the RWR index and the Katz index is superior to the coupling effects between RWR and CN, PA and LP index.Further, the RWR index plays a leading role in the coupling algorithm.Compared with directly setting up a model for network evolution, link prediction is more effective in analyzing the cooperation evolution mechanism of supply chain networks.%应用供应链网络结构或节点的属性信息预测未产生链接的节点企业合作的可能性是链路预测应用供应链网络合作演化分析的关键,利用链路预测的理论框架和评价方法,借助5种相似性指标对能源供应链网络合作连边演化进行预测.研究结果表明:当使用供应链网络结构属性作为单一相似性指标时,RWR指标的预测效果最好;耦合指标预测精确度要比单独考虑单一指标时有显著提高;RWR指标和Katz指标耦合效果要优于和CN指标、PA指标、LP指标耦合效果,且RWR指标在耦合算法中起到主导性作用;与直接建立网络演化模型相比,链路预测在分析供应链网络合作演化机制上更为有效.