Link Prediction in Relational Data

TLDR

Real‑world domains are relational, comprising objects linked in complex ways. The study aims to predict both the presence and type of links between entities in such domains. The authors use a relational Markov network to model the entire link graph with probabilistic subgraph patterns, applying it to university webpage and social network datasets. RMNs with subgraph patterns significantly improve link prediction accuracy compared to flat classification.

Abstract

Many real-world domains are relational in nature, consisting of a set of objects related to each other in complex ways. This paper focuses on predicting the existence and the type of links between entities in such domains. We apply the relational Markov network framework of Taskar et al. to define a joint probabilistic model over the entire link graph — entity attributes and links. The application of the RMN algorithm to this task requires the definition of probabilistic patterns over subgraph structures. We apply this method to two new relational datasets, one involving university webpages, and the other a social network. We show that the collective classification approach of RMNs, and the introduction of subgraph patterns over link labels, provide significant improvements in accuracy over flat classification, which attempts to predict each link in isolation.