社交网络中基于交互行为的影响最大化研究

发布时间：2018-04-04 01:06

  本文选题：社交网络　切入点：影响最大化　出处：《云南大学》2016年硕士论文

【摘要】：近几年来,随着各种社交网络的迅猛发展,人与人之间的主要交流方式逐渐从线下变为线上,这样,就产生了在社交网络中如何查找最有影响力的k个用户的问题,也就是社交网络中影响最大化问题。影响最大化问题就是挖掘社交网络中最有影响力的Top-k个节点集。之前的影响最大化问题研究中,大多只是根据网络的拓扑结构来查找最有影响力的用户,而忽略了反映用户之间亲密程度的一个很重要的因素——交互行为,从而使挖掘出的最有影响力的用户往往与实际情况有较大偏差。基于此种考虑,本文提出了基于交互行为的影响最大化问题,建立了一个基于用户交互行为的影响传播模型UIB_IC模型。在UIB_IC模型中,为了对交互行为的大小进行定量化表示,本文提出了交互度的概念,给出了基于用户交互行为的影响力计算方法,并进行了归一化处理,将之作为用户之间的激活概率。这样,本文就根据UIB_IC模型,提出了GAUIB算法。GAUIB算法是在贪心算法的基础上改进的,它将用户之间的交互行为运用到用户之间能否激活成功的概率中,这样就能够更加准确地衡量用户之间的影响力大小。在GAUIB算法中,因为其具有子模性,所以该算法可以达到63%的准确性。为了提高该算法的计算效率,之后本文又对其进行了优化,使用CELF算法减少了计算量,使其效率有了很大提升。最后,本文通过从腾讯微博中得到的相关数据进行实验验证,证明GAUIB算法可以得到基于用户交互行为的影响最大化用户集合S。
[Abstract]:In recent years, with the rapid development of various social networks, the main way of communication between people has gradually changed from offline to online, thus the question of how to find the most influential k users in social networks has arisen.This is the problem of maximizing influence in social networks.The problem of maximizing influence is mining the most influential Top-k node set in social networks.Most of the previous studies on impact maximization only looked for the most influential users based on the topology of the network, ignoring an important factor that reflected the degree of closeness between users-interaction.As a result, the most influential users excavated often deviate from the actual situation.Based on this consideration, this paper proposes the problem of maximizing the impact based on interaction behavior, and establishes a model of impact propagation based on user interaction behavior (UIB_IC).In the UIB_IC model, in order to quantify the size of the interaction behavior, the concept of interaction degree is proposed, and the influence calculation method based on the user interaction behavior is presented, and the normalized processing is given.Use this as the activation probability between users.Therefore, according to the UIB_IC model, this paper proposes that the GAUIB algorithm. GAUIB algorithm is improved on the basis of greedy algorithm, which applies the interaction behavior between users to the probability of the success of activation between users.This makes it possible to measure the impact between users more accurately.In the GAUIB algorithm, the accuracy of the algorithm can reach 63% because of its submodule.In order to improve the computational efficiency of the algorithm, this paper then optimizes the algorithm, using the CELF algorithm to reduce the amount of calculation, so that its efficiency has been greatly improved.Finally, through the experimental verification of relevant data obtained from Tencent Weibo, it is proved that the GAUIB algorithm can obtain the maximum user set based on user interaction behavior.
【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】：TP393.09
，

本文编号：1707700