社会网络影响力最大化研究及其网络营销应用
发布时间：2018-07-31 14:25
【摘要】：建立合理的社会网络节点(成员)影响力评估体系、识别影响力节点,是分析一个网络组织结构的关键问题,对于研究网络中节点之间的影响力分布乃至传播模式都具有重要意义。对节点影响力的评估不仅是网络舆情引导者决策时的重要信息来源,也是实施网络病毒营销时选择种子节点的重要依据。常见的识别方法是通过比较节点中心性测度数值的大小,来获得最具影响力的节点。然而每一种中心性测度都有自己的优势和劣势。TOPSIS作为一种多属性决策手段已经成为决策的一个重要分支,而将TOPSIS用来识别网络中的影响力节点尚不多见。本文选取具有代表性的多属性决策方法--TOPSIS方法作为主要研究方法,提出基于熵权的TOPSIS拓展方法,并将新方法用于社会网络节点影响力评估的模型中。论文的主要内容有：首先,阐述了论文的研究背景和意义,并对网络营销和社会网络影响力最大化研究的现状进行了文献综述,并提出了论文总的研究思路和框架。然后,论文介绍了网络营销的背景知识、TOPSIS方法的理论基础和具体计算步骤,系统介绍了社会网络影响力评估的常见方法,包括影响力节点识别方法和影响力最大化问题。接着,论文对社会网络影响力做了界定,并分析了影响力的相关因素包括时间、节点位置和话题等,比较了几个常见影响力最大化问题的模型,包括贪心算法、线性阀值模型和独立级联模型等。在介绍了常见网络分析中心性方法后,重点对TOPSIS方法展开了研究和拓展。本文使用熵权法来确定TOPSIS方法中多种属性的权重,提出了基于熵权的TOPSIS拓展方法。基于熵权的TOPSIS法首先对社会网络的多属性,即网络中节点的多种不同中心性测度进行属性权重的确定,再用经典TOPSIS法来集成这些多属性,从而获得每个节点的重要性估值及其排名,达到识别网络中影响力节点的目标。将网络节点不同的中心性测度作为TOPSIS多属性的新方法在一定程度上克服了单一中心性测度方法的局限性和劣势,而熵权法的引入则是增强了TOPSIS方法的客观性。紧接着论文通过对历史文献的比较参考,选取了具有典型意义的社会网络分析中心性指标。最后,我们用Susceptible-Infected (SI)模型来比较新方法和常见的单一中心性测度方法的性能。数据实验结果显示了新方法相比常见方法的高效性和可行性。这将为节点识别研究在舆情控制、病毒营销等方面的应用节约时间成本,从而达到更好的效果。
[Abstract]:Establishing a reasonable social network node (member) influence evaluation system and identifying the influence nodes is the key problem for analyzing a network organization structure. It is of great significance to the study of the distribution of influence and even the mode of communication among nodes in the network. The evaluation of the influence of the nodes is not only the heavy duty of the network public opinion guide. The source of information is also an important basis for the selection of seed nodes in the network virus marketing. The common method of recognition is to obtain the most influential nodes by comparing the size of the node centrality measure value. However, each central measure has its own advantages and disadvantages.TOPSIS as a multi attribute decision-making method. As an important branch of decision making, it is still rare to use TOPSIS to identify the influence nodes in the network. In this paper, a representative multi attribute decision making method --TOPSIS method is selected as the main research method, and the TOPSIS extension method based on entropy weight is proposed, and the new method is used in the model of the social network node influence evaluation. The main contents are as follows: first, it expounds the research background and significance of the paper, and reviews the current status of the research on network marketing and social network influence maximization, and puts forward the general research ideas and framework of the paper. Then, the paper introduces the background knowledge of network marketing, the theoretical basis and concrete calculation step of the TOPSIS method. The common methods of social network impact assessment are introduced, including the identification of influence nodes and the maximization of influence. Then, the thesis defines the influence of social network, and analyzes the factors related to the influence, including time, node position and topic, and compares the models of several common influence maximization problems. Type, including greedy algorithm, linear threshold model and independent cascade model. After introducing the central method of common network analysis, this paper focuses on the research and expansion of the TOPSIS method. In this paper, entropy weight method is used to determine the weight of multiple attributes in the TOPSIS method, and the TOPSIS extension method based on entropy weight is proposed. The entropy weight based TOPSIS head method is proposed. The multiple attributes of the social network, that is, to determine the attribute weights of a variety of different centrality measures of the nodes in the network, and then integrate these multiple attributes with the classical TOPSIS method, obtain the importance valuation and ranking of each node, and achieve the target of identifying the influence nodes in the network. The central measure of the network nodes is taken as the central measure. The new method of TOPSIS multi attributes overcomes the limitation and disadvantage of the single central measure method to a certain extent, and the introduction of entropy weight method is to enhance the objectivity of the TOPSIS method. After the comparison of the historical literature, the paper selects the central index of the social network analysis with typical significance. Finally, we use Sus The ceptible-Infected (SI) model compares the performance of the new method and the common single centrality measurement method. The results of the data experiment show the efficiency and feasibility of the new method compared with the common methods. This will save time and cost for the application of node recognition in public opinion control, virus marketing and so on, thus achieving better results.
【学位授予单位】：东南大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：F224
[Abstract]:Establishing a reasonable social network node (member) influence evaluation system and identifying the influence nodes is the key problem for analyzing a network organization structure. It is of great significance to the study of the distribution of influence and even the mode of communication among nodes in the network. The evaluation of the influence of the nodes is not only the heavy duty of the network public opinion guide. The source of information is also an important basis for the selection of seed nodes in the network virus marketing. The common method of recognition is to obtain the most influential nodes by comparing the size of the node centrality measure value. However, each central measure has its own advantages and disadvantages.TOPSIS as a multi attribute decision-making method. As an important branch of decision making, it is still rare to use TOPSIS to identify the influence nodes in the network. In this paper, a representative multi attribute decision making method --TOPSIS method is selected as the main research method, and the TOPSIS extension method based on entropy weight is proposed, and the new method is used in the model of the social network node influence evaluation. The main contents are as follows: first, it expounds the research background and significance of the paper, and reviews the current status of the research on network marketing and social network influence maximization, and puts forward the general research ideas and framework of the paper. Then, the paper introduces the background knowledge of network marketing, the theoretical basis and concrete calculation step of the TOPSIS method. The common methods of social network impact assessment are introduced, including the identification of influence nodes and the maximization of influence. Then, the thesis defines the influence of social network, and analyzes the factors related to the influence, including time, node position and topic, and compares the models of several common influence maximization problems. Type, including greedy algorithm, linear threshold model and independent cascade model. After introducing the central method of common network analysis, this paper focuses on the research and expansion of the TOPSIS method. In this paper, entropy weight method is used to determine the weight of multiple attributes in the TOPSIS method, and the TOPSIS extension method based on entropy weight is proposed. The entropy weight based TOPSIS head method is proposed. The multiple attributes of the social network, that is, to determine the attribute weights of a variety of different centrality measures of the nodes in the network, and then integrate these multiple attributes with the classical TOPSIS method, obtain the importance valuation and ranking of each node, and achieve the target of identifying the influence nodes in the network. The central measure of the network nodes is taken as the central measure. The new method of TOPSIS multi attributes overcomes the limitation and disadvantage of the single central measure method to a certain extent, and the introduction of entropy weight method is to enhance the objectivity of the TOPSIS method. After the comparison of the historical literature, the paper selects the central index of the social network analysis with typical significance. Finally, we use Sus The ceptible-Infected (SI) model compares the performance of the new method and the common single centrality measurement method. The results of the data experiment show the efficiency and feasibility of the new method compared with the common methods. This will save time and cost for the application of node recognition in public opinion control, virus marketing and so on, thus achieving better results.
【学位授予单位】：东南大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：F224
【相似文献】
相关期刊论文 前10条
1 李国富;汪张林;桂云苗;;基于制造企业核心竞争力的AHP-TOPSIS分析[J];科技创业月刊;2009年10期
2 徐时波;;基于TOPSIS法的省域旅游业综合竞争力分析[J];知识经济;2011年16期
3 王s，
本文编号：2155870
本文链接：https://www.wllwen.com/jingjilunwen/jingjiguanlilunwen/2155870.html
