基于自然生长模型与重构模型的社会网络结构建模与实验分析
发布时间：2018-07-26 14:15
【摘要】：社会网络结构建模是社会网络其它领域研究的基础,旨在构建合理的社会网络结构模型。但由于隐私保护等诸多原因,几乎不可能获取构建社会网络结构模型所需的全部数据,因此研究利用不完整数据构建社会网络结构模型十分必要。现有的构建社会网络结构模型的方法虽然有许多,但都存在一定的问题,比如早期的静态模型并不具有社会网络的全部拓扑特性,大部分的动态模型中节点随时间推进无限增长。本文首先从应用场景分析出发,提炼技术需求,然后分别从时间、空间以及时空融合三个不同角度,设计了三种社会网络结构模型——基于历史事件的自然生长模型、基于局部网络信息的重构模型以及时空融合社会网络模型。基于历史事件的自然生长模型是一种基于行为原则以及事件对网络结构影响机制的动态模型;基于局部网络信息的重构模型建立在修正的共邻相似性和属性相似性两种链路预测指标之上;时空融合社会网络模型的构建则结合了社会网络自然选择等基础理论。基于Netlogo实验平台,本文使用不同的实验数据及方法,对上述三种网络模型进行了实验验证。实验证明,三种网络模型分别在不同条件下具有较好的准确性。因此,具体应用场景下,可以根据可获取数据的不同,选择相应的模型建立社会网络结构。此外,由于外在环境和节点自身状态的不断变化,社会网络的结构也随之不断变化,但社会网络的这种变化总是使其结构在整体上向着更加稳定的方向发展。
[Abstract]:The modeling of social network structure is the basis of other fields of social network research, aiming at constructing a reasonable social network structure model. However, for many reasons, such as privacy protection, it is almost impossible to obtain all the data needed to construct the social network structure model, so it is very necessary to study the use of incomplete data to construct the social network structure model. Although there are many existing methods of constructing social network model, there are some problems. For example, the early static model does not have all the topological characteristics of social network. In most dynamic models, nodes grow infinitely with time. In this paper, first of all, from the perspective of application scenario analysis, we refine the technical requirements, and then from three different angles of time, space and space fusion, we design three kinds of social network structure models-natural growth model based on historical events. Reconstruction model based on local network information and spatiotemporal fusion social network model. The natural growth model based on historical events is a dynamic model based on behavior principle and the influence mechanism of event on network structure. The reconstruction model based on local network information is based on two kinds of link prediction indexes: the modified co-neighbor similarity and attribute similarity, while the space-time fusion social network model combines the basic theories of social network natural selection and so on. Based on the Netlogo experimental platform, this paper uses different experimental data and methods to verify the above three network models. Experiments show that the three network models have better accuracy under different conditions. Therefore, according to the difference of available data, the corresponding model can be selected to establish the social network structure in the specific application scenario. In addition, the structure of social network changes with the change of external environment and the state of node itself, but the change of social network always makes the structure of social network more stable as a whole.
【学位授予单位】：南京邮电大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：O157.5
本文编号：2146277
[Abstract]:The modeling of social network structure is the basis of other fields of social network research, aiming at constructing a reasonable social network structure model. However, for many reasons, such as privacy protection, it is almost impossible to obtain all the data needed to construct the social network structure model, so it is very necessary to study the use of incomplete data to construct the social network structure model. Although there are many existing methods of constructing social network model, there are some problems. For example, the early static model does not have all the topological characteristics of social network. In most dynamic models, nodes grow infinitely with time. In this paper, first of all, from the perspective of application scenario analysis, we refine the technical requirements, and then from three different angles of time, space and space fusion, we design three kinds of social network structure models-natural growth model based on historical events. Reconstruction model based on local network information and spatiotemporal fusion social network model. The natural growth model based on historical events is a dynamic model based on behavior principle and the influence mechanism of event on network structure. The reconstruction model based on local network information is based on two kinds of link prediction indexes: the modified co-neighbor similarity and attribute similarity, while the space-time fusion social network model combines the basic theories of social network natural selection and so on. Based on the Netlogo experimental platform, this paper uses different experimental data and methods to verify the above three network models. Experiments show that the three network models have better accuracy under different conditions. Therefore, according to the difference of available data, the corresponding model can be selected to establish the social network structure in the specific application scenario. In addition, the structure of social network changes with the change of external environment and the state of node itself, but the change of social network always makes the structure of social network more stable as a whole.
【学位授予单位】：南京邮电大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：O157.5
【参考文献】
相关期刊论文 前3条
1 杨波;游新冬;段文奇;;复杂动态网络演化社团结构探测分析的研究进展[J];计算机应用研究;2013年05期
2 方锦清;汪小帆;郑志刚;毕桥;狄增如;李翔;;一门崭新的交叉科学:网络科学(上)[J];物理学进展;2007年03期
3 张文宏,阮丹青,潘允康;天津农村居民的社会网[J];社会学研究;1999年02期
，本文编号：2146277
本文链接：https://www.wllwen.com/kejilunwen/yysx/2146277.html
