具有时间感知的加权网络链路预测研究

发布时间：2018-02-14 22:40

本文关键词： 社会网络链路预测加权网络时间感知　出处：《中南大学》2014年硕士论文　论文类型：学位论文

【摘要】：链路预测是社会网络研究中的一个重要分支,其主要任务是通过当前网络中存在的链路对未来有可能出现的链路进行预测。近年来,社会网络中的链路预测问题吸引了越来越多的研究人员关注,它在现代科学中不仅具有深刻的理论意义,而且具有极大的应用价值。由于现有的链路预测算法主要应用于简单的静态无权网络中,没有考虑权重以及时间序列对于链路预测的影响,所以不能很好地适应复杂网络的情形。为了解决这一问题,本文针对含权网络,提出了节点权重和链路权重的概念,在扩展和结合已有链路预测算法的基础上提出了新的链路预测算法。此外,针对时间感知网络,提出了时间因子的概念,用来量化时间序列中的时间因素以帮助计算节点间的相似度；最后,综合权重思想和时间因子的概念提出了基于时间感知的加权网络链路预测算法。本文针对加权网络和时间感知网络,分别选取了若干个真实数据集对提出的链路预测算法进行了分析比较。实验结果表明,本文所提出的加权链路预测算法相比无权链路预测算法有更好的精度,基于时间感知的链路预测算法比不含时间因子的链路预测算法效果更明显,结合了二者优点的加权时间感知链路预测算法也获得了较高的预测精度。图21幅,表10个,参考文献65篇。
[Abstract]:Link prediction is an important branch of social network research. Its main task is to predict the possible future links through the existing links in the current network. The problem of link prediction in social networks has attracted more and more researchers' attention. It not only has profound theoretical significance in modern science, but also has great application value. Because the existing link prediction algorithms are mainly used in simple static unweighted networks and do not consider the influence of weight and time series on link prediction, they can not adapt well to the situation of complex networks. In this paper, the concepts of node weight and link weight are proposed for weighted networks, and a new link prediction algorithm is proposed on the basis of extending and combining existing link prediction algorithms. The concept of time factor is proposed to quantify the time factor in time series to help calculate the similarity between nodes. Finally, a weighted network link prediction algorithm based on time perception is proposed based on the concept of weight and time factor. In this paper, several real data sets are selected to analyze and compare the proposed link prediction algorithms for weighted networks and time-aware networks. The experimental results show that, The weighted link prediction algorithm proposed in this paper has better precision than the unweighted link prediction algorithm. The link prediction algorithm based on time perception is more effective than the link prediction algorithm without time factor. The weighted time perceptual link prediction algorithm which combines the advantages of the two methods also has a high prediction accuracy. Fig. 21, table 10, references 65.
【学位授予单位】：中南大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP393.09

【参考文献】