What the study found
The study found that research on automatic classification of software issue reports covers multiple techniques, including traditional machine learning, deep learning-based methods, and large language models. It also found that the existing literature has several gaps, especially limited practitioner involvement and a focus on prediction accuracy over other adoption factors.
Why the authors say this matters
The authors say a comprehensive overview can help identify future research directions and collect potentially relevant existing solutions. The study suggests that future work should focus on real industrial evaluations, consider factors such as explainability, scalability, and generalizability, and involve practitioners.
What the researchers tested
The researchers conducted a systematic mapping study on automatic techniques for classifying issue reports into bugs and non-bugs. They identified 46 studies on the topic and reviewed the methods and characteristics reported in that literature.
What worked and what didn't
The literature includes a range of classification techniques, from traditional machine learning to deep learning and large language models. However, the studies mostly relied on archival data from open-source repositories, did not involve practitioners, and mainly evaluated prediction accuracy rather than explainability, scalability, or generalizability.
What to keep in mind
The abstract does not describe limitations of the mapping study itself beyond the observed gaps in the literature. The findings are based on 46 studies and on issue-report classification research as represented in the available literature.
Key points
- The review identified 46 studies on automatic issue report classification.
- Existing studies use traditional machine learning, deep learning, and large language models.
- The literature mainly focuses on prediction accuracy rather than explainability, scalability, or generalizability.
- Practitioners are not widely involved in the studied work.
- Most studies rely on archival data from open-source repositories.
Disclosure
- Research title:
- Automatic issue report classification uses multiple machine learning approaches
- Image credit:
- Photo by Tima Miroshnichenko on Pexels
Get the weekly research newsletter
Stay current with peer-reviewed research without reading academic papers — one filtered digest, every Friday.