大型社交网络社区结构演化

Evolution of Communities in Large Social Network

大型社交网络已经成为互联网最主要的组成部分,是人们获取信息、分享交流的主要渠道。而其中的社区结构指的是社交网络中一些人呈现出的紧紧聚集的群落关系,同一社区内的用户往往拥有相同的兴趣话题。以往对社区结构的研究大多集中于使用无监督的社区发现算法在大型社交网络中给出用户的社区划分方法。而针对社交网络中社区对应的拓扑结构,重点在时间维度上考察以社区结构为基础的邻接图的固有特征对其社区成长的影响,利用有监督的机器学习方法,给出各个特征的重要性排名以及预测社区成员增长率的预测模型。研究数据集主要基于豆瓣小组功能。

As a main part of Internet, large social network has become the main platform for people to gain information and share their thoughts in recent years. Community in social network indicates the group of users who are connected densely and share same topics and interests. Past works focus on the unsupervised learning algorithm of exploring the potential community structures While this paper studies the structure of communities which are already labeled in social network, and gives both a prediction model to predict the community growth and a rank of feature importance. The data are built on Douban group.