结合节点度和节点聚类系数的链路预测算法

发布时间：2018-07-08 16:45

本文选题：复杂网络 + 链路预测　；参考：《小型微型计算机系统》2017年07期

【摘要】：链路预测作为复杂网络的一个重要研究方向,基于节点相似性指标进行预测是最为常用的一种方法.传统的链路预测方法通常使用共同邻居数目或节点的度来衡量节点之间的相似性.节点对之间的关系不仅与邻居节点数目和度有关,节点的聚类系数体现了节点的聚集能力,对产生链接会起到一定的作用.基于这个观点,提出一种结合节点度和聚类系数的链路预测算法.利用共同邻居节点的度和聚类系数计算被预测节点对之间的相似性.不仅充分利用网络局部结构信息,还能够体现出共同邻居节点之间的差异性.在十组实际数据集上的实验结果表明,提出的链路预测算法与传统的五个算法(CN,AA,RA,PA,Jaccard)和基于聚类系数的CCLP算法相比具有很好的预测效果.
[Abstract]:As an important research direction of complex networks, link prediction based on node similarity index is the most commonly used method. Traditional link prediction methods usually use the number of common neighbors or the degree of nodes to measure the similarity between nodes. The relationship between node pairs is not only related to the number and degree of neighbor nodes, the clustering coefficient of nodes reflects the clustering ability of nodes, and it will play a certain role in producing links. Based on this viewpoint, a link prediction algorithm combining node degree and clustering coefficient is proposed. The degree and clustering coefficient of common neighbor nodes are used to calculate the similarity between predicted pairs of nodes. Not only make full use of the network local structure information, but also reflect the differences between the common neighbor nodes. Experimental results on ten sets of actual data sets show that the proposed link prediction algorithm has a good prediction effect compared with the traditional five algorithms (CNA ARAA PAP Jaccard) and the CCLP algorithm based on clustering coefficients.
【作者单位】：安徽大学计算机科学与技术学院;
【基金】：国家自然科学基金项目(61175046、61402006)资助安徽省自然科学基金项目(1508085MF113)资助教育部人文社科基金项目(14YJC860020)资助
【分类号】：O157.5

【相似文献】