Improving Relational Classification Using Link Prediction Techniques

  • Cristina Pérez-Solà
  • Jordi Herrera-Joancomartí
Conference paper

DOI: 10.1007/978-3-642-40988-2_38

Volume 8188 of the book series Lecture Notes in Computer Science (LNCS)
Cite this paper as:
Pérez-Solà C., Herrera-Joancomartí J. (2013) Improving Relational Classification Using Link Prediction Techniques. In: Blockeel H., Kersting K., Nijssen S., Železný F. (eds) Machine Learning and Knowledge Discovery in Databases. ECML PKDD 2013. Lecture Notes in Computer Science, vol 8188. Springer, Berlin, Heidelberg

Abstract

In this paper, we address the problem of classifying entities belonging to networked datasets. We show that assortativity is positively correlated with classification performance and how we are able to improve classification accuracy by increasing the assortativity of the network. Our method to increase assortativity is based on modifying the weights of the edges using a scoring function. We evaluate the ability of different functions to serve for this purpose. Experimental results show that, for the appropriated functions, classification on networks with modified weights outperforms the classification using the original weights.

Preview

Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

Copyright information

© Springer-Verlag Berlin Heidelberg 2013

Authors and Affiliations

  • Cristina Pérez-Solà
    • 1
  • Jordi Herrera-Joancomartí
    • 1
    • 2
  1. 1.Dept. d’Enginyeria de la Informació i les ComunicacionsUniversitat Autònoma de BarcelonaBellaterraSpain
  2. 2.Internet Interdisciplinary Institute (IN3)UOCSpain