基于网络资源流量的链路预测方法

在复杂网络中，现有基于结构相似性的链路预测方法较少考虑全局和局部拓扑信息之间平衡性、准确度和复杂度之间平衡性以及网络资源动态流动的问题。将网络资源流量作为相似性判断依据，提出一种准局部链路预测方法。根据网络中节点重要性的不同来为它们分配对应的资源，以保证资源分配的合理性。针对网络资源提出一种动态流动机制，将节点对双向流动的资源之和作为相似程度的量化指标。引入节点对之间中间路径节点的概念，分析中间路径节点在资源流动过程中的稀释作用。在此基础上，计算初始资源量和稀释作用量从而得到网络资源流量方法的性能评估指标值。在Jazz、NS等11个真实世界的网络中进行实验，对比该方法与CN、Salton等常见基准方法在准确度和鲁棒性方面的性能表现，结果表明，所提方法能够充分利用准局部信息，既能考虑资源流动性又能解决平衡性问题，可有效提高链路预测性能。

Link Prediction Method Based on Network Resource Traffic

In complex networks, the existing link prediction methods based on structural similarity rarely consider the balance between global and local topology information, balance between accuracy and complexity, or dynamic flow of network resources.Therefore, taking the network resource traffic as the basis of similarity judgment, a quasi-local link prediction method is proposed.According to the different importance of nodes in the network, corresponding resources are allocated to them to ensure a rationality of resource allocation.A dynamic flow mechanism for network resources is proposed, which takes the sum of the two-way flow resources of nodes as a quantitative index of similarity.The concept of intermediate path nodes between node pairs is introduced, and the dilution effect of intermediate path nodes during resource flow is analyzed.On this basis, the initial resource and dilution effects are calculated to obtain the performance evaluation index value of the network resource traffic method.Experiments are conducted with eleven real-world networks including Jazz and NS, and the performance of the proposed method is compared with common benchmark methods such as CN and Salton in terms of accuracy and robustness.The results show that the proposed method can fully use quasi-local information, consider resource flow, solve balance problems, and effectively improve the link-prediction performance.