Neighbor-based Link Prediction with Edge Uncertainty
- Chi Zhang
- Osmar R. Zaiane, University of Alberta (Database)
In this work, we are concerned with uncertain networks and focus on the problem of link prediction with edge uncertainty. Networks with edge uncertainty are networks where connections between nodes are observed with some probability. We propose the uncertain version of the popular neighbors-based metrics for link prediction. The metrics are developed by considering all possible worlds generated by the uncertain network. We state that by taking all possible worlds of the uncertain network into account, the performance of link prediction can be improved. Since uncertain edges result in a very large number of possible worlds, we propose an efficient divide and conquer algorithm to reduce time complexity and calculate these metrics. Finally, we evaluate our metrics using existing ground truth to show the effectiveness of our proposed approach against other popular link prediction methods.
Citation
C. Zhang, O. Zaiane. "Neighbor-based Link Prediction with Edge Uncertainty". Pacific-Asia Conference on Knowledge Discovery and Data Mining (PAKDD), Macau, China, pp 462-474, April 2019.| Keywords: | social network analysis, link prediction, uncertain networks |
| Category: | In Conference |
| Web Links: | Webdocs |
BibTeX
@incollection{Zhang+Zaiane:PAKDD19,
author = {Chi Zhang and Osmar R. Zaiane},
title = {Neighbor-based Link Prediction with Edge Uncertainty},
Pages = {462-474},
booktitle = {Pacific-Asia Conference on Knowledge Discovery and Data Mining
(PAKDD)},
year = 2019,
}Last Updated: September 15, 2020Submitted by Sabina P
