Systematic Biases in Link Prediction: comparing heuristic and graph embedding based methods

Aakash Sinha 1 Rémy Cazabet 2, 3, 4, 5 Rémi Vaudaine 4
1 IIT Delhi - Indian Institute of Technology Delhi
2 DM2L - Data Mining and Machine Learning
LIRIS - Laboratoire d'InfoRmatique en Image et Systèmes d'information
3 IXXI - Institut Rhône-Alpin des systèmes complexes
4 LIRIS - Laboratoire d'InfoRmatique en Image et Systèmes d'information
5 UCBL - Université Claude Bernard Lyon 1
Abstract : Link prediction is a popular research topic in network analysis. In the last few years, new techniques based on graph embedding have emerged as a powerful alternative to heuristics. In this article, we study the problem of systematic biases in the prediction, and show that some methods based on graph embedding offer less biased results than those based on heuristics, despite reaching lower scores according to usual quality scores. We discuss the relevance of this finding in the context of the filter bubble problem and the algorithmic fairness of recommender systems.
Keywords : link prediction systematic biases filter bubble graph embedding
Document type :
Conference papers
Complete list of metadatas

Cited literature [26 references]  Display  Hide  Download
https://hal-udl.archives-ouvertes.fr/hal-01892997
Contributor : Remy Cazabet <>
Submitted on : Monday, October 29, 2018 - 7:45:27 PM
Last modification on : Thursday, February 7, 2019 - 4:54:33 PM

Files

Author_start_from_here.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : hal-01892997, version 2
  • ARXIV : 1811.12159

Collections

UNIV-LYON2 | UNIV-LYON3 | EC-LYON | ENS-LYON | UNIV-LYON1 | INSA-LYON | INRIA | LIRIS | INSA-GROUPE | UGA

Citation

Aakash Sinha, Rémy Cazabet, Rémi Vaudaine. Systematic Biases in Link Prediction: comparing heuristic and graph embedding based methods. Complex networks 2018 - The 7th International Conference on Complex Networks and Their Applications, 2018, Cambridge, United Kingdom. ⟨hal-01892997v2⟩

Export

BibTeX TEI DC DCterms EndNote

Share

Metrics

Record views

133

Files downloads

161

Contact

CCSD