基于模体演化的多因子动态链路预测方法  

MULTI FACTOR DYNAMIC LINK PREDICTION METHOD BASED ON MOTIF EVOLUTION

作　　者：赵宇红[1] 张晓炜 Zhao Yuhong;Zhang Xiaowei(School of Information Engineering,Inner Mongolia University of Science and Technology,Baotou 014010,Inner Mongolia,China)

机构地区：[1]内蒙古科技大学信息工程学院,内蒙古包头014010

出　　处：《计算机应用与软件》2022年第3期234-240,共7页Computer Applications and Software

基　　金：国家自然科学基金项目(61562065)。

摘　　要：为了提高动态网络链路预测准确率,从网络结构微观演化角度,提出基于模体演化的多因子动态链路预测方法(MFME)。在动态网络时间窗口划分优化的基础上,引入整合移动平均自回归模型构建预测模体演化的概率矩阵,综合考虑模体演化影响因子及模体演化概率,可获得任意节点间的连接边概率。在真实数据集的实验表明,所提方法能达到更好的链路预测效果。To make the accuracy of dynamic link prediction higher,from the perspective of micro-evolution of network structure,a multi factor dynamic link predication method(MFME)is proposed.Based on the optimization of dynamic network time window size division,the integrated moving average autoregressive model was introduced to construct the probability matrix for predicting the motif evolution.Considering the influence factors and probability of motif evolution,the continuous edge probability between any nodes could be obtained.The experiments on real data set show that the proposed method can achieve better link prediction effect.

关 键 词：动态链路预测 模体演化 时间窗口 整合移动平均自回归模型 模体演化影响因子 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]