Link prediction in bipartite graphs using internal links and weighted projection

Oussama Allali,Clemence Magnien,Matthieu Latapy

doi:10.1109/infcomw.2011.5928947

Oussama Allali, Clemence Magnien + Show 1 more

Open Access

https://doi.org/10.1109/infcomw.2011.5928947

Copy DOI

Abstract

Many real-world complex networks, like client-product or file-provider relations, have a bipartite nature and evolve during time. Predicting links that will appear in them is one of the main approach to understand their dynamics. Only few works address the bipartite case, though, despite its high practical interest and the specific challenges it raises. We define in this paper the notion of internal links in bipartite graphs and propose a link prediction method based on them. We describe the method and experimentally compare it to a basic collaborative filtering approach. We present results obtained for two typical practical cases. We reach the conclusion that our method performs very well, and that internal links play an important role in bipartite graphs and their dynamics.

Highlights

Many real-world complex networks have a natural bipartite structure and may be modeled as bipartite graphs [1], i.e. two sets of nodes with links only between nodes in different sets
We study the performance of our method on two real-world datasets. We show that this method reaches very good performances and that internal links play a key role in the dynamics of real-world bipartite graphs
We show that the amount of internal links in them is high, which ensures the relevance of predicting internal links

Summary

INTRODUCTION

Many real-world complex networks have a natural bipartite structure and may be modeled as bipartite graphs [1], i.e. two sets of nodes with links only between nodes in different sets. Typical examples include peer-to-peer file-provide graphs [2] where peers are linked to the files they provided; and clientproduct graphs where clients are linked to the products they bought [3]. Most of these networks are dynamic: they evolve during time, with node and link additions and removals. We address here the problem of link prediction in bipartite graphs. We study the performance of our method on two real-world datasets We show that this method reaches very good performances and that internal links play a key role in the dynamics of real-world bipartite graphs.

RELATED WORK

THE BIPARTITE LINK PREDICTION PROBLEM

INTERNAL LINK PREDICTION

EXPERIMENTAL RESULTS

Weight functions

Impact of the number of predicted links

Impact of the weight function

CONCLUSION