基于DynamicRank的重要节点集挖掘算法

为了更为有效地挖掘复杂网络中综合影响力高的节点集、提高重要节点集挖掘算法的准确性，针对节点集中各节点在信息传播中不同程度的相互促进和相互抑制作用，基于DynamicRank算法设计了一种新的级联概率计算模型。通过增强贪心策略，优先从种子集邻居以外的节点中选取备选节点，减小种子集内重叠邻居对信息传播引发的抑制作用；利用级联概率计算模型，增强种子集对信息传播的级联促进作用。在4个实际网络上采用SIR模型进行了测试，实验结果表明，相较于已有重要节点挖掘方法H-index、Degree、DynamicRank、VoteRank和EnRenew，提出的算法挖掘出的节点集具有更高影响力。特别地，在Grid数据集上，本文方法挖掘出的种子集的传播范围比对比算法平均提升了49.3%。提出的算法针对信息传播概率和贪心策略的改进有利于重要节点集的挖掘和识别。

Critical nodes mining algorithm based on DynamicRank

In order to mine the node sets with high comprehensive influence in complex networks effectively and to improve the accuracy of critical nodes mining algorithm, we design a novel cascade probability model based on DynamicRank algorithm according to the different degrees of promotion and suppression of node sets in information propagation. By enhancing the greedy strategy, the candidate nodes are preferentially selected from the nodes outside the seed set neighbors to reduce the inhibition of overlapping neighbors in the seed set on information transmission. The cascading probability calculation model is used to enhance the cascading promotion effect of seed set on information transmission. The performance of proposed model is evaluated on four real networks under SIR spreading model. Experimental results show that the seeds mined by our method have higher influence than those by benchmark methods such as H-index, Degree, DynamicRank, VoteRank and EnRenew. In particular, on the Grid dataset, compared with other critical nodes mining methods, the propagation scale from initial spreaders mined by our method is improved by 49.3% on average. It demonstrates that the improvement on information propagation probability and greedy strategies is conducive to mining and identifying critical nodes.