Quantifying Community Growth in Dynamic Social Networks

Community detection in social networks is a method that helps us to discover groups of users that are tightly connected. So far, most research has focused on detecting communities in static networks. However, networks evolve over time, and so do the communities within these networks. Understanding the evolution of communities is important to gain insight into macroscopic trends that drive changes in the network. Such an understanding may also help us to predict future growth and trends within the network. In this paper we will present a new model and algorithm for identifying changing, or dynamic, communities in evolving networks. We use will use dynamic communities to develop the concept of community growth rate, which quantifies how fast a community grows relative to the overall network. We then show that the community growth rate is only one example of a family of more general community evolution metrics. Finally, we evaluate our model on real-world data using the Angellist dataset, a network of entrepreneurs, and show how the community evolution event we identified make intuitive sense based on our understanding of the domain.