 |
International Society of Science and Applied Technologies |
| ADM: An Agile Development Manager for Early Defect Prediction in Small Software Teams | ||||
| Author |
Kazuhira Okumoto
|
|||
| Co-Author(s) |
|
|||
| Abstract | Small, agile, and CI/CD software teams operate in
short cycles, with sparse data and limited quality assurance resources, making early quality assessment and defect prediction particularly difficult. Traditional reliability and defect models, developed for larger, more stable projects, often fail to provide timely or reliable insights in these environments. This paper introduces the Agile Development Manager (ADM), a lightweight analytics module that extends the STAR quality-intelligence platform to small-scale projects. ADM applies segmented linear modeling over short rolling windows to capture evolving defect trends and provide early, stability-aware predictions with minimal data requirements. Experimental results using real project data show that ADM enables a 46% reduction in the time required to achieve stable prediction. This result demonstrates that ADM not only accelerates stabilization in absolute days but also significantly reduces the proportion of the lifecycle required to obtain reliable predictive confidence. ADM supports early risk visibility, improves prediction behavior, and enables proactive corrective actions in fast-cycle development environments.
|
|||
| Keywords | Agile software analytics, defect prediction, lightweight reliability modeling, CI/CD quality management, software quality intelligence | |||
| Article #: RQD2026-35 | ||||
Proceedings of 31st ISSAT International Conference on Reliability & Quality in Design |