A Dynamic Model Selection Approach to Mitigate the Change of Balance Problem in Cross-Version Bug Prediction

Hiroshi Demanou, Akito Monden, Masateru Tsunoda

Research output: Contribution to journalConference articlepeer-review

Abstract

This paper focuses on the “change of balance” problem in cross-version bug prediction where the percentage of buggy modules changes between different versions. Such difference badly affects the prediction performance. To mitigate this problem, this paper employs a dynamic model selection approach equipped with two prediction models (always-buggy model and always-non-buggy model) and Bandit algorithm to select better models in each one-module-by-one prediction. An experiment with data sets of 20 releases of 10 open source software showed that the proposed approach can improve F1-measure compared with the conventional cross-version prediction.

Original languageEnglish
Pages (from-to)4-9
Number of pages6
JournalCEUR Workshop Proceedings
Volume3330
Publication statusPublished - 2022
EventJoint of the 10th International Workshop on Quantitative Approaches to Software Quality and the 6th Software Engineering Education Workshop, QuASoQ-SEED 2022 - Virtual, Online
Duration: Dec 6 2022 → …

Keywords

  • Bandit algorithm
  • defect-prone module prediction
  • software quality assurance

ASJC Scopus subject areas

  • Computer Science(all)

Fingerprint

Dive into the research topics of 'A Dynamic Model Selection Approach to Mitigate the Change of Balance Problem in Cross-Version Bug Prediction'. Together they form a unique fingerprint.

Cite this