面向主题耦合的影响力最大化研究

发布时间：2018-01-24 13:23

  本文关键词： 社会网络 影响力最大化 藕合相似度 主题　出处：《云南大学》2016年硕士论文　论文类型：学位论文

【摘要】：随着网络逐渐成为人与人之间主要的社交方式,在网络中挖掘最有影响力的用户成为非常值得关注的问题。商品通过网络进行营销已经成为商业战场中主流的进攻方式,利用好网络中用户自身的影响效应是取得事半功倍效果的关键所在。如此一来,社交网络中的影响力最大化问题便成为研究的焦点。影响力最大化问题就是要在社交网络中确定有限个种子节点,使得这些用户能够在网络中引起最大的影响效应。关于该问题的研究已经有很多成熟的理论,比如影响力最大化问题中经典的贪心算法。但是传统的影响力最大化问题并没有考虑到网络中传播的信息具有的不同主题以及这些主题之间的关系,这在一定程度上局限了影响力最大化问题的求解精度。本文在影响力最大化问题的基础上提出了面向主题耦合的影响力最大化问题,并且针对该问题提出了GACT (Greedy Algorithm based on the Couped Topics)算法来挖掘在特定的传播主题下最具有影响力的用户。GACT算法首先分析网络中不同主题之间的耦合相似性,而后使用潜在语义索引的方法计算用户对于不同主题的偏好,在综合考虑主题之间耦合相似性与用户对不同主题偏好的基础上扩展独立级联模型,在扩展的传播模型上使用经典的贪心算法挖掘最具有影响力的用户,最后使用CELF算法进行优化以提高算法的时间效率。与经典的影响力最大化算法相比,GACT算法能够考虑到传播主题之间的耦合相似性并且能够与用户偏好更为有效的结合,在影响力最大化问题中挖掘出更为精确的种子节点。最后,在电影社交网络上通过实验证明了GACT算法相比经典的影响力最大化算法能够更为有效的挖掘在特定主题下最具有影响力的用户。
[Abstract]:As the Internet has gradually become the main form of social interaction between people. Mining the most influential users in the network has become a matter of great concern. Commodity marketing through the network has become the mainstream attack way in the business battlefield. Making good use of the influence of the user in the network is the key to achieve twice the result with half the effort. The problem of maximization of influence in social networks becomes the focus of research. The problem of maximization of influence is to determine a limited number of seed nodes in social networks. So that these users can cause the greatest impact in the network. There are many mature theories about this problem. For example, the classical greedy algorithm in the influence maximization problem, but the traditional impact maximization problem does not take into account the different topics and the relationship between the information spread in the network. To a certain extent, it limits the accuracy of solving the problem of maximization of influence. Based on the problem of maximization of influence, this paper puts forward the problem of maximization of influence oriented to subject coupling. To solve this problem, GACT greedy Algorithm based on the Couped Topics is proposed. Algorithm to mine the most influential user. Gas algorithm under a specific transmission topic. Firstly, the coupling similarity between different topics in the network is analyzed. Then the potential semantic index is used to calculate user preferences for different topics, and the independent cascading model is extended based on the consideration of the coupling similarity between topics and user preferences for different topics. In the extended propagation model, the classical greedy algorithm is used to mine the most influential users. Finally, the CELF algorithm is used to optimize to improve the time efficiency of the algorithm. GACT algorithm can take into account the coupling similarity between propagating topics, and can be combined with user preferences more effectively, mining more accurate seed nodes in the problem of maximizing the impact. Finally. The experiments on the film social network show that the GACT algorithm is more effective than the classical influence maximization algorithm in mining the most influential users under a particular topic.
【学位授予单位】：云南大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP301.6;F274
，

本文编号：1460144