Is the commit buggy? – Catalin Meirosu, Ericsson – Product Development Group Software – Defined Infrastructure & Abgeiba Yaroslava Isunza Navarro, Modulai AB


Identifying defective code (bugs) in commits from a software repository may improve productivity and reduce the amount of corrections that need to be applied in production. We present a framework based on deep learning that could help identify buggy commits. The algorithms are trained on a well-known cloud infrastructure management code repository

Key Takeaways

  • Deep learning could be used for identifying buggy commits
  • Accuracy of identification differs depending on the approach
  • Open source dataset and example code are available to facilitate reproduction and building applications around the framework

Add comment