Toward a better scientific collaboration success prediction model through the feature space expansion

Fahimeh Ghasemian, Kamran Zamanifar*, Nasser Ghasem-Aqaee, Noshir Contractor

*Corresponding author for this work

Research output: Contribution to journalArticle

8 Scopus citations

Abstract

The problem with the prediction of scientific collaboration success based on the previous collaboration of scholars using machine learning techniques is addressed in this study. As the exploitation of collaboration network is essential in collaborator discovery systems, in this article an attempt is made to understand how to exploit the information embedded in collaboration networks. We benefit the link structure among the scholars and also among the scholars and the concepts to extract set of features that are correlated with the collaboration success and increase the prediction performance. The effect of considering other aggregate methods in addition to average and maximum, for computing the collaboration features based on the feature of the members is examined as well. A dataset extracted from Northwestern University’s SciVal Expert is used for evaluating the proposed approach. The results demonstrate the capability of the proposed collaboration features in order to increase the prediction performance in combination with the widely-used features like h-index and average citation counts. Consequently, the introduced features are appropriate to incorporate in collaborator discovery systems.

Original languageEnglish (US)
Pages (from-to)777-801
Number of pages25
JournalScientometrics
Volume108
Issue number2
DOIs
StatePublished - Aug 1 2016

Keywords

  • Collaboration network
  • Collaboration success
  • Collaborator discovery system
  • Hypergraph
  • Machine learning
  • Scientific collaboration

ASJC Scopus subject areas

  • Social Sciences(all)
  • Computer Science Applications
  • Library and Information Sciences

Fingerprint Dive into the research topics of 'Toward a better scientific collaboration success prediction model through the feature space expansion'. Together they form a unique fingerprint.

  • Cite this