基于符号网络和动态网络社区检测的研究
发布时间：2018-12-15 13:52
【摘要】：复杂网络系统存在于我们生活中,影响着我们的日常生活。绝大多数的网络都具有聚类特性,都存在社区结构,如何快速在真实的网络系统中发现社区结构已成为社会关注的热点之一。针对此种网络社区检测问题,本文从两个方面入手:(1)符号网络社区检测问题:本文提出一种基于亲密度动态演化的符号网络社区检测算法。加入了新的相似度计算函数,并为了使不连接的两节点之间有相似度,加入了最短路径的相似度计算函数。构造符号网络节点间亲密度演化的动力学模型,从而使同一个社区中节点的亲密度随着时间的变化更新为1,不同社区之间节点的亲密度随着时间的变化更新为0。在本文所提出网络模型的基础上,整个网络会分为几个不同的社区。为了验证算法的性能,本文针对USC真实网络,GGS真实网络以及17个人工合成网络进行了仿真,并与已有文献作了相关比较,实验结果表明算法有一定的优势。(2)动态网络社区检测问题:本文在符号网络社区检测算法的基础上,加入了时间的因素。首先对网络中上一个时刻的相似度和下一个时刻的相似度进行了加权,其次在Attractor模型上进行了改进,使得亲密度高的节点在一个社区,最后把网络分为不同的社区,并通过真实网络和人工合成网络检测了算法的有效性。为了验证加权系数α对该算法的影响,我们对α取了不同的值进行比较,发现α的取值对算法并没有影响。
[Abstract]:Complex network system exists in our life and affects our daily life. Most networks have clustering characteristics and community structure. How to quickly find community structure in real network system has become one of the hot spots in the society. Aiming at this kind of network community detection problem, this paper starts with two aspects: (1) symbol network community detection problem: this paper proposes a symbol network community detection algorithm based on dynamic evolution of affinity density. A new similarity calculation function is added and the shortest path similarity calculation function is added in order to make the two unconnected nodes have similarity. The dynamic model of the evolution of the affinity between nodes in the symbolic network is constructed, so that the affinity of the nodes in the same community changes with time to 1, and that of the nodes between different communities changes to 0. Based on the proposed network model, the whole network will be divided into several different communities. In order to verify the performance of the algorithm, this paper simulates the USC real network, GGS real network and 17 artificial synthetic networks. The experimental results show that the algorithm has some advantages. (2) the dynamic network community detection problem: this paper adds the time factor to the symbolic network community detection algorithm. Firstly, the similarity between the last moment and the next moment in the network is weighted. Secondly, the Attractor model is improved to make the nodes with high affinity in one community. Finally, the network is divided into different communities. The validity of the algorithm is verified by real network and artificial synthetic network. In order to verify the influence of the weighting coefficient 伪 on the algorithm, we compare the different values of 伪 and find that the value of 伪 has no effect on the algorithm.
【学位授予单位】：内蒙古工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：O157.5
[Abstract]:Complex network system exists in our life and affects our daily life. Most networks have clustering characteristics and community structure. How to quickly find community structure in real network system has become one of the hot spots in the society. Aiming at this kind of network community detection problem, this paper starts with two aspects: (1) symbol network community detection problem: this paper proposes a symbol network community detection algorithm based on dynamic evolution of affinity density. A new similarity calculation function is added and the shortest path similarity calculation function is added in order to make the two unconnected nodes have similarity. The dynamic model of the evolution of the affinity between nodes in the symbolic network is constructed, so that the affinity of the nodes in the same community changes with time to 1, and that of the nodes between different communities changes to 0. Based on the proposed network model, the whole network will be divided into several different communities. In order to verify the performance of the algorithm, this paper simulates the USC real network, GGS real network and 17 artificial synthetic networks. The experimental results show that the algorithm has some advantages. (2) the dynamic network community detection problem: this paper adds the time factor to the symbolic network community detection algorithm. Firstly, the similarity between the last moment and the next moment in the network is weighted. Secondly, the Attractor model is improved to make the nodes with high affinity in one community. Finally, the network is divided into different communities. The validity of the algorithm is verified by real network and artificial synthetic network. In order to verify the influence of the weighting coefficient 伪 on the algorithm, we compare the different values of 伪 and find that the value of 伪 has no effect on the algorithm.
【学位授予单位】：内蒙古工业大学
【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 陈建芮;洪志敏;汪丽娜;乌兰;;Dynamic evolutionary community detection algorithms based on the modularity matrix[J];Chinese Physics B;2014年11期
2 程苏琦;沈华伟;张国清;程学旗;;符号网络研究综述[J];软件学报;2014年01期
3 何东晓;周栩;王佐;周春光;王U，
本文编号：2380768
本文链接：https://www.wllwen.com/kejilunwen/yysx/2380768.html
