Big Data Driven Similarity Based U-Model for Online Social Networks

Abstract

The proliferation of information technologies results in a complex network evolution of online social networks. Traditional model driven aided description cannot be appropriate for the dynamic evolution of social networks. However, in this paper, relying on the big data collected from a range of real-world online social networks, we try to explore the underlying evolution for online social networks. Firstly, we define a pair of big data driven similarity based utility models (U- models), i.e. the undirected U-model as well as the directed U-model, which can effectively reflect the statistical characteristics of online social networks. Secondly, we analyze the small-world property, scale-free property and high clustering coefficient property of our proposed U-models which consider nodes' similarity, popularity and asymmetry in a network. Finally, relying on three real-world big datasets, i.e. Sina Weibo, Tencent Weibo and Twitter, sufficient experiments show that the U-models outperform the traditional models in portraying the evolution statistical characteristic of online social networks.