New perspectives and methods in link prediction

Abstract

This paper examines important factors for link prediction in networks and provides a general, high-performance framework for the prediction task. Link prediction in sparse networks presents a significant challenge due to the inherent disproportion of links that can form to links that do form. Previous research has typically approached this as an unsupervised problem. While this is not the first work to explore supervised learning, many factors significant in influencing and guiding classification remain unexplored. In this paper, we consider these factors by first motivating the use of a supervised framework through a careful investigation of issues such as network observational period, generality of existing methods, variance reduction, topological causes and degrees of imbalance, and sampling approaches. We also present an effective flow-based predicting algorithm, offer formal bounds on imbalance in sparse network link prediction, and employ an evaluation method appropriate for the observed imbalance. Our careful consideration of the above issues ultimately leads to a completely general framework that outperforms unsupervised link prediction methods by more than 30% AUC.

RolX: Role Extraction and Mining in Large Networks

“…Features have been extracted from graphs for several data mining tasks. In [15], the authors propose extracting topological features for pairs of nodes for link prediction. In [16], the authors develop a multi-level framework to detect anomalies in time-varying graphs based on graph, sub-graph, and node level features.…”

Proximity measures for link prediction based on temporal events

“…Initially, four classical proximity scores based on neighborhood were considered: PA, CN, AA and JC. These are very popular measures used in the state of the art of link prediction (Bringmann et al, 2010;Lichtenwalter et al, 2010). Each measure is explained below.…”

“…As previously mentioned, the starting point of this approach is to extract the values/scores of different measures that indicate the similarity between pairs of nodes. These scores can be used either by unsupervised (Liben-Nowell et al, 2003;Lu & Zhou, 2011;Murata & Moriyasu, 2008) or supervised link prediction (Hasan et al, 2006;Lichtenwalter, Lussier, & Chawla, 2010;Sá & Prudêncio, 2011). In the former approach, a proximity measure is chosen and deployed to rank node pairs in the network.…”

Link Prediction in Heterogeneous Collaboration Networks

“…A variety of techniques for addressing this problem have been explored including graph theory, metric learning, statistical relational learning, matrix factorization, and probabilistic graphical models [1,[15][16][17]. This chapter is an extended version of our prior work on supervised link prediction models [9].…”