An Adaptive Boosting Ensemble Pipeline for Software Defect Prediction over Imbalanced Dataset
IOSR Journal of Computer Engineering · 2026-05 · 28 (5) · 32-46
DOI: 10.9790/0661-2805013246 ↗Kurz gefasst
Fachartikel in der Zeitschrift IOSR Journal of Computer Engineering, 2026-05. Verfasst von Stephen Bassi Joseph und 2 weiteren Autor:innen.
Abstract (Originalsprache)
Background
Software Defect Prediction (SDP) is a proactive quality assurance strategy that uses static code complexity metrics to catch software bugs before a system is deployed. software metrics are highly imbalanced. Because clean code segments vastly outnumber defective ones, standard machine learning models tend to ignore the minority class, leading to a dangerous bias that leaves actual defects undetected. This study develops a reliable, end-to-end machine learning pipeline engineered to neutralize class imbalance, eliminate featurescaling distortions, and establish a dependable balance between precision and recall. Materials and Methods: The proposed workflow processes the NASA CM1 dataset by enforcing strict out-ofsample data binarization and Z-score normalization to prevent data leakage. We then deploy an adaptive ensemble framework that links ten deep decision tree base learners under the discrete SAMME AdaBoost algorithm.
Results
After implementing the proposed method on an independent 33% test data, the ensemble pipeline achieves an average overall accuracy of 99.40%. This shows the veracities of the dataset’s skewness, the model isolates defective modules with a localized precision of 0.98 and a recall of 0.98, while maintaining a stable macro-averaged 𝐹1 -score of 0.97 and a weighted 𝐹1 -score of 0.99 across all code streams.
Conclusion
The overall results show that combining strict training-boundary and normalization with sequential error correction vectors prevent minority-class blindness in the model. Without using noisy artificial data oversampling, this method enables the ensemble engine to map complex, non-linear code trajectories and identify irregularities spontaneously
Zitieren
APA
Joseph, S. B., Mustapha, K., & Ismail, A. A. (2026). An Adaptive Boosting Ensemble Pipeline for Software Defect Prediction over Imbalanced Dataset. IOSR Journal of Computer Engineering, 28(5), 32-46. https://doi.org/10.9790/0661-2805013246
Vancouver
Joseph SB, Mustapha K, Ismail AA. An Adaptive Boosting Ensemble Pipeline for Software Defect Prediction over Imbalanced Dataset. IOSR Journal of Computer Engineering. 2026;28(5):32-46. doi:10.9790/0661-2805013246
BibTeX
@article{Joseph2026,
author = {Joseph, Stephen Bassi and Mustapha, Kunduli and Ismail, Abideen Adekunle},
title = {An Adaptive Boosting Ensemble Pipeline for Software Defect Prediction over Imbalanced Dataset},
journal = {IOSR Journal of Computer Engineering},
year = {2026},
volume = {28},
number = {5},
pages = {32--46},
doi = {10.9790/0661-2805013246}
}
EUVIMED