Using hierarchies in online social networks to determine link prediction

Abstract

Social hierarchy is one of the most basic concepts in sociology and analysis of networks. In today’s world, online social networks are very popular, and thus, there is a need to analyze social hierarchies in the form of complex networks. Social hierarchies can help in improving link prediction, optimize page rank, provide efficient query results, etc. These networks can provide us with important information about social interactions and how to efficiently use it. Social network will be modeled as directed graph, and then, social hierarchies are explored by converting directed graph to directed acyclic graph. By converting to directed acyclic graph (DAG), we can partition the graph into different levels which will divide the network into multiple hierarchies. The topmost level signifies highest social ranking, while bottom ones represent the least ranking. Most algorithms widely used calculate DAG in O(n3), which is fine for small graphs, but when dealing with large social networks, it is not practically feasible. Thus here, we devise a new method to compute DAG. In this method, we will remove all the cycles from graph G′. In addition to this, we use a two-phase algorithm—compute and refine. The result is used to improve link prediction. The algorithm used two orders faster than the basic algorithm.