基于共同邻居惩罚的复杂网络链路预测方法

Complex Network Link Prediction Method Based on Common Neighbor Punishment

摘要:

链接预测是确定用户间关系的基本工具。通过相似性度量进行链路预测是一种常见的方法，提出一种基于相似度的链路预测算法，根据网络结构及拓扑特性来确定相似度，引入优化链路预测度量方法，将聚类系数作为网络结构性质。此外，并考虑共享邻域，得到较其他同类链路预测方法更好的性能。实验结果表明，提出的算法性能优于经典算法。结合在Facebook、Twitter与新浪微博等社交网络环境中的实验结果可知，SLP-CNP法较其他算法具有更优精度与效率。在未来的工作中，还可尝试在所提方法的基础上，提升在加权网络、有向网络和二部网络中的适用性。

Abstract:

Link prediction is a fundamental tool for determining relationships between users. Link prediction by similarity measure is a common method. This paper proposes a link prediction algorithm based on similarity, which determines the similarity according to the network structure and topological characteristics, introduces the optimized link prediction measure, and takes the clustering coefficient as the network structure property. In addition, considering the shared neighborhood, the performance is better than other similar link prediction methods. Experimental results show that the proposed algorithm outperforms the classical algorithm. Combined with the experimental results in the social network environment such as Facebook, Twitter and Sina Weibo, it can be seen that SLP-CNP method has better accuracy and efficiency than other algorithms. In the future work, we can also try to improve the applicability of the proposed method in weighted networks, directed networks and bipartite networks.