技术交易网络社团结构检测方法研究与实证

发布时间：2018-06-05 18:37

本文选题：复杂网络 + 技术交易网络　；参考：《北京邮电大学》2017年硕士论文

【摘要】：国内外学者已经成功的将复杂性科学的一些思想应用于刻画和解释技术交易网络,但是这些尝试主要停留在宏观层面,针对技术交易网络中观及微观层面的定性定量研究还少见报道。相比而言,通过对网络中社团结构的深入细致分析,可以更全面深刻的掌握系统结构本身存在的规律,有利于构建良好的技术交易模式,促进技术交易活动的顺利进行。本文将复杂网络社团结构检测算法应用到技术交易行为的研究过程当中,以实际交易数据拟合网络生成模型,通过技术交易网络对社团检测算法进行实证与分析。主要展开以下方面的具体工作:(1)技术交易网络模型的构建。在构建出复杂技术交易网络之后,提取网络静态特性及动态特性,对网络结构进行初步理解。(2)分析多种社团检测方法并进行实证。分别对层次聚类算法、模块度优化算法、谱聚类算法以及流传播算法四类社团检测算法进行分析。通过对比模块度、时间复杂度等社团划分评价标准发现:模块度优化类算法对于技术交易网络的切合程度较高。(3)深入分析技术交易网络中重要节点及典型社团,发掘网络内部结构规律。首先,对多种中心性进行加权获得新的中心性评估标准;其次,通过重要节点所在社团的演化研究发现:社团内部主要通过重要节点吸引外部节点融入到社团当中。通过对网络内部社团聚类后的结果发现,节点度大的社团向外分枝较多,节点度小的社团内联系较为紧密。本文通过研究与实证相结合,以北京市技术交易网络为基础进行社团划分,发现使用模块度优化类算法在时间复杂度及划分结果等方面均具有较好表现,且通过算法发现的社团内部结构规律能够在一定程度上为对技术交易市场的管理提供指导与支持。
[Abstract]:Scholars at home and abroad have successfully applied some ideas of complexity science to the description and interpretation of technology trading networks, but these attempts are mainly at the macro level. Qualitative and quantitative studies on the meso-and micro-level of technology trading networks are rarely reported. In contrast, through the thorough and careful analysis of the community structure in the network, we can grasp the rules of the system structure more comprehensively and profoundly, which is conducive to the construction of a good technology trading model and the smooth progress of the technology trading activities. In this paper, the complex network community structure detection algorithm is applied to the research process of technical transaction behavior. The network generation model is fitted with the actual transaction data, and the community detection algorithm is empirically analyzed through the technology transaction network. The main work of the following aspects of the specific work: 1) the construction of technology trading network model. After the complex technical transaction network is constructed, the static and dynamic characteristics of the network are extracted, and the network structure is preliminarily understood. The hierarchical clustering algorithm, modular optimization algorithm, spectral clustering algorithm and flow propagation algorithm are analyzed respectively. By comparing the evaluation criteria of community division, such as modularity and time complexity, it is found that the modularity optimization algorithm has a higher degree of relevance to the technical trading network. The important nodes and typical communities in the technology trading network are analyzed in depth. Explore the internal structure of the network law. Firstly, new evaluation criteria of centrality are obtained by weighting various kinds of centrality. Secondly, through the study of the evolution of the community in which the important nodes are located, it is found that the internal community mainly attracts the external nodes into the community through the important nodes. The results show that the communities with large nodes are more branched out and the communities with small nodes are more closely connected. Through the combination of research and demonstration, this paper divides communities based on Beijing technology trading network, and finds that the modular degree optimization algorithm has better performance in terms of time complexity and partition results. And the internal structure of the community found by the algorithm can provide guidance and support for the management of the technology trading market to a certain extent.
【学位授予单位】：北京邮电大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：O157.5

【参考文献】