复合网的链路预测

发布时间：2018-04-18 11:40

本文选题：复杂网络 + 复合网　；参考：《青岛大学》2017年硕士论文

【摘要】：近年,复杂网络技术兴起,链路预测作为复杂网络研究热点也被广泛应用到诸多领域。链路预测分为考虑节点属性信息的局部方法和使用网络结构信息的全局方法:局部方法计算简便,但是存在数据难以获取的缺点;全局方法因考虑全面,精度较高,但是计算复杂性也较高。本文重点研究半局部方法,基于复合网的链路预测对于致病基因预测和霾波动幅度规律两个方面进行了研究。本文涉及的主要工作有:(1)针对传统链路预测方法计算节点相似性时忽略节点间相互作用不对称性,导致相同权值和路长但权重分布不同的节点间的相似性难以区分的问题。本文基于Itan的相似性指标,提出了相互作用非对称指标——IA(Interaction Asymmetry)指标,并在视网膜色素变异病(RP)数据集上进行了实验验证,取得了较好的预测效果,证明了该指标的有效性。(2)为研究霾的发生受哪些因素的影响以及其影响机理,基于山东省3年多的空气质量检测数据,对山东省内的霾波动规律展开了研究。本文建立了应用于霾问题研究的多子网复合复杂网络,提出了基于复合网的时序链路预测模型,为与气象以时间为关联做分析提供基础。
[Abstract]:In recent years, with the rise of complex network technology, link prediction as a research hotspot of complex networks has been widely used in many fields.Link prediction can be divided into local method considering node attribute information and global method using network structure information. The local method is simple to calculate, but it has the shortcoming of difficult to obtain data.But the computational complexity is also high.In this paper, the semi-local method is mainly studied, and the link prediction based on compound network is studied in two aspects: the prediction of pathogenic gene and the regularity of haze fluctuation amplitude.The main work involved in this paper is: (1) for the traditional link prediction method, the asymmetry of the interaction between nodes is ignored when calculating the node similarity, which leads to the difficulty of distinguishing the similarity between nodes with the same weight and different weight distribution.Based on the similarity index of Itan, this paper puts forward the interaction asymmetric index, IAP interaction metric index, and carries out the experimental verification on the data set of retinal pigment mutation disease, and obtains the better prediction effect.It is proved that the effectiveness of this index. (2) in order to study the influence of haze occurrence and its mechanism, based on the air quality detection data of more than 3 years in Shandong Province, the regularity of haze fluctuation in Shandong Province has been studied.In this paper, a multi-subnet complex network applied to haze research is established, and a time-series link prediction model based on composite network is proposed, which provides a basis for the time-dependent analysis with meteorology.
【学位授予单位】：青岛大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：O157.5

【参考文献】