基于Motif的社交网络用户影响力排序方法研究

摘要

LeaderRank已被广泛地用于衡量社交网络中用户的影响力或重要性的排序算法，但传统的LeaderRank仅利用基于边的关系，而忽略由少量节点组成的子图捕获的高阶关系。在本文中，我们提出一种基于Motif的LeaderRank (MLR)算法，将Motif的高阶关系合并到LeaderRank算法中，提高社交网络用户影响力的排序效果。我们在Twitter数据集上进行实验，结果不仅表明MLR算法的可行性，并且还显著提高了在社交网络用户影响力排序的准确性。除了与基线算法比较之外，还对MLR算法的参数进行分析，表明在社交网络中对用户影响力排序时，MLR算法比LeaderRank要更好。 LeaderRank has been widely used in ranking algorithms to measure the influence or importance of users in social networks, but traditional LeaderRank only uses edge-based relationships and ignores high-order relationships captured by a subgraph composed of a small number of nodes. In this paper, we propose a Motif-based LeaderRank (MLR) algorithm, which incorporates the higher-order relationships of Motif into the LeaderRank algorithm to improve the ranking effect of social network user influence. We perform experiments on the Twitter dataset, and the results not only show the feasibility of the MLR algorithm, but also significantly improve the accuracy of the ranking of user influence in social networks. In addition to comparing with the baseline algorithm, the parameters of the MLR algorithm are also analyzed, which shows that the MLR algorithm is better than LeaderRank when ranking user influence in social networks.