Introduction
Link prediction in bipartite graphs is a problem arising in a variety of problem context, most prominently
Explaining the science
Due to the survivor bias dominating the observed data in the available tendering data, we are interested in
Bipartite graphs arise in a wide range of problems but methodologies for their analysis are in their infancy compared
Project aims
- Understand sensitivity to link sparsity, for existing techniques for community detection and link prediction.
- Develop a novel approach to community detection that addresses the particular case of bipartite graph structures
and overcomes the limitations of traditional projection-based approaches to community-detection.
- Develop a novel link prediction method for bipartite graphs that allows for the frictionless integration of a range of
additional data sources including information in text format, through the integration of multiple data views.
Applications
The graph-based, link prediction algorithms will be integrated into a recommender system modelled on the
- Guiding suppliers and buyers to relevant contracts.
- Increasing transparency of public contracts resulting in better value for tax-payers.
- Increasing competition which leads to market efficiency and lower prices.
Recent updates
March 2021: Completion of generator
April 2021: First manuscript under submission to CompleNet Live
Organisers
Dr Luis Ospina
Presidential Fellow, University of ManchesterDr Rudy Arthur
Lecturer in Data Science, University of ExeterDr Adrian M. Stetco
Research AssociateContact info
Arielle Bennett
[email protected]