摘要
针对复杂网络链路预测受到网络性质的影响,从而降低复杂网络链路预测效果,提出基于混合深度学习的复杂网络链路预测方法。利用复杂网络在运行过程中的链路变化情况,综合考虑复杂网络历史信息对链路的影响系数,得到了RA指数、AA指数和CN指数等相似指数,基于混合深度学习的反向传播流程,分析复杂网络链路隐藏层的状态,利用复杂网络隐藏层的输出,预测出复杂网络链路输出值,将相似性指标作为复杂网络链路预测的训练样本,构建复杂网络链路预测模型,利用模式分类方法实现多个网络节点之间的链路预测。实验结果表明,基于混合深度学习的复杂网络链路预测方法将时间窗口设为360秒和180秒、样本维度为500和600时,预测效果是最好的,且预测精度较传统方法的预测精度高。
In view of the influence of network properties on complex network link prediction, which reduces the effect of complex network link prediction, a complex network link prediction method based on hybrid deep learning is proposed. Based on the link changes during the operation of the complex network and the influence coefficient of the historical information of the complex network on the link, RA index, AA index and CN index were obtained. According to the back propagation process of hybrid deep learning, the state of link hidden layer in complex networks was investigated in detail. The output of complex network hidden layer predicted the output value of complex network link. Similarity index was adopted as training sample to construct complex network link prediction model. The pattern classification method was applied to achieve the link prediction between multiple network nodes. The experimental results show that when the time window is set to 360 seconds and 180 seconds and the sample dimensions are 500 and 600, the prediction effect is the best, and the prediction accuracy is higher than that of the traditional method.
作者
白雪
董德森
BAI Xue;DONG De-sen(Information Centre,Jilin Institute of Chemical Technology,Jilin Jilin 132022,China;Normal College,Yanbian University,Yanji Jilin 133002,China)
出处
《计算机仿真》
北大核心
2021年第11期309-313,共5页
Computer Simulation
基金
吉林省高教科研课题(JGJX2020C62)
吉林化工学院科学技术研究项目(吉化院合字[2018]第061号)。
关键词
混合深度学习
复杂网络
网络链路
预测模型
Mixed deep learning
Complex network
Network link
Prediction model
作者简介
白雪(1990-),女(朝鲜族),吉林省吉林人,硕士,实验师,主要从事计算机实践教学、现代教育技术方面研究;董德森(1966-),男(汉族),吉林延吉人,硕士,副教授,硕士生导师,主要从事计算机实践教学、信息技术教育方面研究。
