基于频繁闭图关联规则的AS级Internet链路预测方法

目前大多数链路预测方法都是针对丢失链路的结构性预测,缺乏针对未来时刻网络链路的时序性预测,为此提出了一种基于频繁闭图关联规则的链路预测方法。将形式化后的动态网络划分为训练集和测试集,基于Apriori思想从训练集中提取频繁闭图,并根据频繁闭图的时间间隔建立时延分布矩阵,用于表征频繁闭图之间的时序关联规则,在此基础上预测测试集中的网络结构。将该方法运用于不同时间尺度下的AS级Internet动态网络中,结果表明,该方法能够以很高的精确率预测波动型动态网络的链路。

Link Prediction of AS Level Internet Based on Association Rule of Frequent Closed Graphs

The existing link prediction methods are mostly focused on structure link prediction like missing links,but few are about temporal link prediction according to unknown links in future,therefore a link prediction method based on association rules of frequent closed graphs was proposed.Dynamic networks are divided into training set and test test,and frequent closed subgraphs are extracted from training set based on Apriori algorithm,thus time-lag distribution matrix is built to represent the temporal association rules between frequent closed graphs,and then the structure in test set is predicted.The link prediction method was used in the dynamic networks of AS level Internet at different time scales,and experimental results show that this method can efficiently predict links in wavery dynamic networks with high precision.