Link prediction on evolving graphs using matrix and tensor factorizations.
- Turkish National Research Institute of Electronics and Cryptology
The data in many disciplines such as social networks, web analysis, etc. is link-based, and the link structure can be exploited for many different data mining tasks. In this paper, we consider the problem of temporal link prediction: Given link data for time periods 1 through T, can we predict the links in time period T + 1? Specifically, we look at bipartite graphs changing over time and consider matrix- and tensor-based methods for predicting links. We present a weight-based method for collapsing multi-year data into a single matrix. We show how the well-known Katz method for link prediction can be extended to bipartite graphs and, moreover, approximated in a scalable way using a truncated singular value decomposition. Using a CANDECOMP/PARAFAC tensor decomposition of the data, we illustrate the usefulness of exploiting the natural three-dimensional structure of temporal link data. Through several numerical experiments, we demonstrate that both matrix- and tensor-based techniques are effective for temporal link prediction despite the inherent difficulty of the problem.
- Research Organization:
- Sandia National Laboratories (SNL), Albuquerque, NM, and Livermore, CA (United States)
- Sponsoring Organization:
- USDOE
- DOE Contract Number:
- AC04-94AL85000
- OSTI ID:
- 1021699
- Report Number(s):
- SAND2010-3823C; TRN: US201117%%291
- Resource Relation:
- Conference: Proposed for presentation at the BIT 50 %3CU%2B2013%3E Trends in Numerical Computing held June 17-20, 2011 in Lund, Sweden.
- Country of Publication:
- United States
- Language:
- English
Similar Records
Scalable tensor factorizations with missing data.
CPOPT : optimization for fitting CANDECOMP/PARAFAC models.