位置感知影响最大化算法及传播模型设计与实现
发布时间：2018-11-24 11:01
【摘要】：影响力最大化问题首先被Domingos和Richardson引入到社会网络领域,成为社会网络领域的一个热门的研究问题。问题提出后各领域的学者纷纷开始提出各种各样的算法用于求解社会网络上的影响力最大化问题。本文针对于具有地理位置信息的商店进行影响最大化问题的研究,主要的研究内容如下:基于喜好及位置因素的贪心算法的研究。现有的边概率取值上一成不变的使用没有现实意义的选取方式,本文通过提出基于喜好相似度与距离两个因素来定义边概率和根据距离来选取候选种子集合的方式,通过这样的方式不仅具有现实的意义而且使图中的无效节点在开始时就被排除。提出基于影响成功模型的区域划分算法,在以往的影响最大化算法中,都是在整个网络上去获取种子节点,这样会花费很大的开销。本算法是通过对位置感知网络划分区域,对每个区域进行种子节点的选取,把各区域得到的节点集合起来得到最终的种子节点集合。本文又提出一个基于影响成功率的传播模型,该模型是考虑到当节点影响其他节点时会根据之前成功激活节点的个数为其自身赋予一个影响成功率,影响成功率会影响之后它的邻居节点是否被激活。最后,本文使用一部分真实数据和一部分模拟数据进行实验验证,并从时间和影响力两个方面对基于喜好及位置因素的贪心算法和基于影响成功率模型的区域划分算法进行验证。
[Abstract]:The problem of maximization of influence is first introduced into the field of social network by Domingos and Richardson, which has become a hot research problem in the field of social network. Since the problem was put forward, scholars in various fields have begun to put forward a variety of algorithms to solve the problem of maximization of influence on social networks. The main contents of this paper are as follows: greedy algorithm based on preferences and location factors. In this paper, we propose a method to define edge probability based on preference similarity and distance and to select candidate seed set according to distance. In this way, not only does it have practical significance, but also the invalid nodes in the graph are excluded from the beginning. A region partition algorithm based on the influence success model is proposed. In the previous impact maximization algorithm, the seed nodes are obtained on the whole network, which will cost a lot of money. The algorithm divides the region of the location-aware network and selects the seed nodes for each region, and then gathers the nodes from each region to get the final seed node set. In this paper, a propagation model based on influence success rate is proposed. This model considers that when nodes affect other nodes, they are given a success rate according to the number of previously successfully activated nodes. The success rate affects whether its neighbor node is activated or not. Finally, some real data and some simulated data are used to verify the experiment. The greedy algorithm based on preference and location factors and the region partition algorithm based on influence success rate model are verified from two aspects of time and influence.
【学位授予单位】：黑龙江大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：O157.5;TP301.6
[Abstract]:The problem of maximization of influence is first introduced into the field of social network by Domingos and Richardson, which has become a hot research problem in the field of social network. Since the problem was put forward, scholars in various fields have begun to put forward a variety of algorithms to solve the problem of maximization of influence on social networks. The main contents of this paper are as follows: greedy algorithm based on preferences and location factors. In this paper, we propose a method to define edge probability based on preference similarity and distance and to select candidate seed set according to distance. In this way, not only does it have practical significance, but also the invalid nodes in the graph are excluded from the beginning. A region partition algorithm based on the influence success model is proposed. In the previous impact maximization algorithm, the seed nodes are obtained on the whole network, which will cost a lot of money. The algorithm divides the region of the location-aware network and selects the seed nodes for each region, and then gathers the nodes from each region to get the final seed node set. In this paper, a propagation model based on influence success rate is proposed. This model considers that when nodes affect other nodes, they are given a success rate according to the number of previously successfully activated nodes. The success rate affects whether its neighbor node is activated or not. Finally, some real data and some simulated data are used to verify the experiment. The greedy algorithm based on preference and location factors and the region partition algorithm based on influence success rate model are verified from two aspects of time and influence.
【学位授予单位】：黑龙江大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：O157.5;TP301.6
【相似文献】
相关期刊论文 前10条
1 闫强;舒华英;;因特网上有害信息传播模型研究[J];管理工程学报;2007年02期
2 冀进朝;韩笑;王U，
本文编号：2353434
本文链接：https://www.wllwen.com/kejilunwen/yysx/2353434.html
