Validation of network measures as indicators of defective modules in software systems

  • Authors:
  • Ayşe Tosun;Burak Turhan;Ayşe Bener

  • Affiliations:
  • Boğaziçi University, Istanbul, Turkey;National Research Council, Ottawa, Canada;Boğaziçi University, Istanbul, Turkey

  • Venue:
  • PROMISE '09 Proceedings of the 5th International Conference on Predictor Models in Software Engineering
  • Year:
  • 2009

Quantified Score

Hi-index 0.01

Visualization

Abstract

In ICSE'08, Zimmermann and Nagappan show that network measures derived from dependency graphs are able to identify critical binaries of a complex system that are missed by complexity metrics. The system used in their analysis is a Windows product. In this study, we conduct additional experiments on public data to reproduce and validate their results. We use complexity and network metrics from five additional systems. We examine three small scale embedded software and two versions of Eclipse to compare defect prediction performance of these metrics. We select two different granularity levels to perform our experiments: function-level and source file-level. In our experiments, we observe that network measures are important indicators of defective modules for large and complex systems, whereas they do not have significant effects on small scale projects.