学术网络主题影响力最大化研究

发布时间：2019-01-01 08:37

【摘要】：随着互联网技术的不断发展和社会网络的流行,越来越多研究者被社会网络所吸引,其中影响力最大化问题成为了研究者们持续关注的重点。影响力最大化问题的目标是挖掘社会网络中最有影响力的传播节点,即从社会网络中选取k个初始传播节点,从这些节点开始对整个社会网络进行影响力传播,直到影响节点数量达到最大。影响力最大化问题在市场营销,商品推荐,谣言控制等领域引起了大量研究者的重视,并且具有广泛的应用前景。针对影响力最大化问题,研究者们已经获得了一些成果,例如传统的影响力传播模型和近似最优解的影响力最大化算法。然而传统影响力传播模型仅考虑到社会网络的拓扑结构,而没有考虑到节点与主题的偏好关系,这与真实社会网络相矛盾,而且在很大程度上影响了影响力最大化问题的结果。本文在传统的影响力最大化问题的基础上提出了学术网络主题影响力最大化问题。首先根据作者、论文、词之间的关系建立超图；接着通过张量分解划分主题；然后提出一种新的基于主题因子图的主题激活概率计算方法,最后基于三种不同的影响力传播模型得到三种学术网络主题影响力传播模型：基于信息流模型的主题影响力传播模型、基于MIA模型的主题影响力传播模型和基于线性影响力模型的主题影响力传播模型以及相应的主题影响力最大化算法。通过主题影响力最大化算法,可以得到学术网络中对特定主题影响力最大的论文集。与传统的影响力传播模型相比,学术网络主题影响力传播模型考虑了主题在影响力传播过程的影响。通过在真实学术网络中的验证分析,可以看到学术网络主题影响力传播模型能够利用学术网络中节点之间的主题相关性,有效挖掘到学术网络中对特定主题的影响力最大的论文集。
[Abstract]:With the continuous development of Internet technology and the popularity of social networks, more and more researchers are attracted by social networks. The goal of the impact maximization problem is to excavate the most influential communication nodes in the social network, that is, select k initial communication nodes from the social network, and begin to spread the influence of the whole social network from these nodes. Until the number of affected nodes reaches the maximum. The problem of maximization of influence has attracted a lot of researchers' attention in the fields of marketing, commodity recommendation, rumor control and so on, and has a wide application prospect. For the problem of maximizing the influence, researchers have obtained some achievements, such as the traditional influence propagation model and the influence maximization algorithm which approximates the optimal solution. However, the traditional influence propagation model only considers the topological structure of social network, but does not take into account the preference relationship between nodes and topics, which contradicts the real social network, and to a large extent affects the results of the problem of maximization of influence. Based on the traditional problem of maximizing influence, this paper puts forward the problem of maximizing the influence of academic network theme. Firstly, the hypergraph is established according to the relationship between the author, the thesis, the words, and then the theme is divided by Zhang Liang. Then a new method based on topic factor graph is proposed to calculate the probability of topic activation. Finally, based on three different influence propagation models, three kinds of academic network topic influence propagation models are obtained: the topic influence propagation model based on the information flow model. The theme influence propagation model based on MIA model, the theme influence propagation model based on linear influence model and the corresponding algorithm of theme influence maximization. Through the algorithm of maximization of topic influence, we can get the most influential papers on a particular topic in academic network. Compared with the traditional influence communication model, the influence of topic in the process of influence communication is considered. Through the validation analysis in the real academic network, we can see that the academic network topic influence propagation model can utilize the topic correlation between the nodes in the academic network. Effectively tap into the academic network of the greatest impact on specific topics collection.
【学位授予单位】：浙江大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TP391.1

【相似文献】