含时二分网络中重要节点的识别研究

发布时间：2018-05-29 22:04

本文选题：二分网络 + 重要节点　；参考：《深圳大学》2017年硕士论文

【摘要】：近年来,网络科学的跨越式发展给我们带来了观察现实世界的新视角。真实世界中的许多错综复杂的关系都可以抽象为仅含有简单元素的网络,并利用网络科学理论定量地描述及分析这些系统的性质。而在网络科学领域,对重要节点识别和研究,对于复杂网络的鲁棒性和可靠性的提氋有很重要的实际意义。不同于网络中占多数的一般节点,网络中的重要节点数量不多,但对于网络的结构或功能有着至关重要的作用。同时基于网络的重要节点识别与网络的控制、同步、传播、预测等都有着紧密的联系。由于网络重要节点识别重大的理论研究意义与广泛的实际应用价值,关于重要节点识别方法的研究一直是网络科学领域热门研究方向。本文基于时间维度研究复杂网络下重要节点识别,研究的主要工作内容如下:(1)针对现有二分网络下经典的重要节点识别方法进行研究,详细介绍了IR、CR、IARR、BiHITS、QR等算法,并比较各种方法的优缺点。(2)基于时间维度研究各个排名算法,提出了两个从时间维度量化分析排名结果的指标:新颖性和均衡性。并在后续实验中发现经典的重要节点识别算法存在排名偏向旧节点的问题。(3)针对无权二分网络提出了一种基于H指数的重要节点识别算法,该算法同时考虑了节点权重和时间演化过程,相比传统方法时间新颖性更好且对重要节点的识别准确性更高。在二个实际网络进行测试,发现新算法发现的重要节点新颖性更好。(4)针对加权二分网络提出了基于时间均衡的BiHITS改进算法,该算法在一般迭代算法的基础上增加一个再均衡过程,使得不同时间分组的对象能在一起比较。算法保留原有算法健壮性的基础上大幅提升算法的时间均衡性。实验结果显示新算法提升对新的优秀商品的识别准确性。
[Abstract]:In recent years, the leapfrog development of network science has brought us a new perspective to observe the real world. Many complicated relationships in the real world can be abstracted into networks containing only simple elements, and the properties of these systems can be quantitatively described and analyzed by using the theory of network science. In the field of network science, the identification and research of important nodes is of great practical significance for the enhancement of robustness and reliability of complex networks. Different from the most common nodes in the network, the number of important nodes in the network is not large, but it plays an important role in the structure or function of the network. At the same time, important node recognition based on network is closely related to network control, synchronization, propagation, prediction and so on. Because of the great theoretical research significance and extensive practical application value of network important node recognition, the research on important node recognition method has been a hot research direction in network science field. In this paper, based on time dimension, important node recognition in complex networks is studied. The main work of the study is as follows: (1) the classical important node recognition methods in existing binary networks are studied, and the algorithms such as IRI CRIARRN BiHITSQR and so on are introduced in detail. And compare the advantages and disadvantages of various methods. 2) based on the time dimension to study the various ranking algorithms, proposed two quantitative analysis from the time dimension of the results of the index: novelty and equilibrium. In the subsequent experiments, it is found that the classical important node recognition algorithm has the problem of ranking biased to the old node. (3) an important node recognition algorithm based on H exponent is proposed for the weighted dichotomy network. The algorithm takes into account both the weight of nodes and the evolution process of time. Compared with the traditional method, the algorithm is more novel in time and more accurate in the recognition of important nodes. By testing in two real networks, it is found that the new algorithm has better novelty of important nodes. (4) an improved BiHITS algorithm based on time equalization is proposed for weighted binary networks. This algorithm adds a reequalization process to the general iterative algorithm, which enables the objects of different time groups to be compared together. On the basis of preserving the robustness of the original algorithm, the time equalization of the algorithm is greatly improved. The experimental results show that the new algorithm improves the accuracy of the identification of new excellent goods.
【学位授予单位】：深圳大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：O157.5

【相似文献】