Loading article...
Loading article...
Stream/Area:
-
Subject:
-
Paper ID:
IJSREM63373
Volume:
10 |
5 |
May 2026
Pages:
1-6
Publication Frequency:
Monthly (12 issues per year)
License:International Journal of Scientific Research in Engineering and Management follows the Creative Commons Attribution 4.0 (CC BY 4.0) License, allowing free use and distribution with proper attribution.
Computer Science & Engineering
Jain (Deemed-to-be University)
Computer Science & Engineering
Jain (Deemed-to-be University)
Think of a software engineering team as a manufacturing line where a fraction of the units leaving the line are secretly defective. Testing is the inspection process — but inspection is expensive, and exhaustively inspecting every module is rarely feasible. Software defect prediction (SDP) asks whether a statistical model trained on historical defect records can assign risk scores to new modules, concentrating inspection effort where it will do the most good. This survey examines how that idea has been pursued using machine learning (ML) over two decades. We trace the progression from rudimentary probabilistic classifiers applied to a handful of size metrics to large pre-trained code language models capable of reasoning about defect risk with no labelled examples at all. Core topics include benchmark datasets, feature engineering, classical ML, ensemble methods, deep learning architectures, transformer-based models, cross-project generalisation, class imbalance, evaluation methodology, explainability, and emerging research directions. The survey is intended as both an entry point for researchers new to the area and a structured reference for practitioners integrating ML-driven defect screening into continuous integration pipelines.
KEYWORDS
Two decades of research have transformed software defect prediction from a niche academic curiosity into a mature discipline with a clear performance trajectory. The journey from Naive Bayes counting lines of code [3] to GraphCodeBERT attending to data-flow edges across millions of pre-trained tokens [29] is not merely a story of better algorithms — it is a story of progressively deeper understanding of what information in a software artefact actually predicts fault risk. No algorithm has achieved universal dominance [21]; pre-trained LLMs lead on performance but at interpretability and computational cost [4], [29]; cross-project generalisation remains hard [30], [31]; imbalance handling is non-negotiable [38]; and explainability tools are now mature enough for production deployment [44], [46]. The field's most important unsolved problem is not algorithmic: we do not yet know, with scientific rigour, whether deploying a defect prediction tool in a real development team actually reduces post-release fault rates.
[1] CISQ, "The Cost of Poor Software Quality in the US: A 2020 Report," Consortium for IT Software Quality, Tech. Rep., 2021. [2] N. Fenton and M. Neil, "A critique of software defect prediction models," IEEE Trans. Softw. Eng., vol. 25, no. 5, pp. 675–689, 1999. [3] T. Menzies, J. Greenwald, and A. Frank, "Data mining static code attributes to learn defect predictors," IEEE Trans. Softw. Eng., vol. 33, no. 1, pp. 2–13, 2007. [4] Z. Feng et al., "CodeBERT: A pre-trained model for programming and natural languages," in Findings of EMNLP, 2020, pp. 1536–1547. [5] H. Ding et al., "Large language models are few-shot testers," in Proc. ICSE, 2024. [6] M. Jureczko and L. Madeyski, "Towards identifying software project clusters with regard to defect prediction," in Proc. PROMISE, 2010, pp. 9–18. [7] R. Just, D. Jalali, and M. D. Ernst, "Defects4J: A database of existing faults," in Proc. ISSTA, 2014, pp. 437–440. [8] A. E. Hassan, "The road ahead for mining software repositories," in Proc. FOSE, 2008, pp. 48–57. [9] IEEE, "IEEE Standard Classification for Software Anomalies," IEEE Std 1044-2009, 2010. [10] T. J. Ostrand, E. J. Weyuker, and R. M. Bell, "Predicting the location and number of faults in large software systems," IEEE Trans. Softw. Eng., vol. 31, no. 4, pp. 340–355, 2005.
Publication Date
5/17/2026
Peer Review Type
Double-Blind Peer Review
Review Rounds: 2
Documents
Article Views
0
Total views since publication
PDF Downloads
3
Total downloads since publication
Geographic Reach
5
Digital Preservation: This article is permanently archived on Zenodo and complies with all data preservation standards across major indexing platforms.
Plagiarism Screening
This manuscript has been checked for plagiarism using TurnitinPlagiarismChecker software.
Conflict of Interest
The authors declare no conflict of interest related to this research work.
Funding Information
No funding information available for this research.
Related research in the same subject area
Loading similar articles...