EUVIMEDEUVIMED entdecken
Crossref

Predicting Task Delays in Distributed Agile Software Development Using Support Vector Machines

Rajender Radharam

International Journal of Agile Software Engineering · 2026 · 1 (1) · 1-12

DOI: 10.55124/ijas.v1i1.105 ↗

Kurz gefasst

Fachartikel in der Zeitschrift International Journal of Agile Software Engineering, 2026. Verfasst von Rajender Radharam.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

In the current software development environment, 'Distributed Agile Software Development' has been widely recognized as the most dominant paradigm. This paradigm combines the flexibility of Agile-based software development methodologies with the scalability of geographically dispersed teams. However, this has been coupled with several challenges, particularly with regards to communication, coordination, and management, leading to delays in the software development process. In spite of the wide acceptance of this paradigm, the absence of effective predictive mechanisms that are capable of forecasting these delays is an issue in the domain of software development. This study analyses the existing gap in the field by proposing a framework that uses Machine Learning techniques to predict task delays that are likely to occur in distributed Agile environments. The importance of this research is its capacity to provide signals that can be used by the project managers to take corrective actions at the right time in order to optimize the projects they manage. An artificial dataset with 500 data samples was created to simulate the environment in which the distributed Agile teams operate by considering two aspects of software development tasks. In the case of the predictive model, the RBF kernel was used with the Support Vector Machine classifier. Here, the parameters considered are the size of the teams, the number of commits made, the comments made on the issues, the differences in the time zones, the complexity of the tasks, and the sprint duration. The target variable considered was the task delay. The model performance was checked using various parameters such as Accuracy, Precision, F1-score,Recall, and AUC values, as well as SHAP values. The model accuracy is found to be 99%, and the AUC value is found to be 0.998. Moreover, experiments are validated using the '10-fold cross-validation' method, which confirms the mean accuracy level of the proposed model as 0.978. The SHAP analysis also revealed that task complexity and time zone differences are the major factors that cause delays in the Agile projects. This validates the importance of using the SVM model to understand the nonlinear coordination dynamics of distributed Agile projects. Keywords: Distributed Agile Development, Task Delay Prediction, Support Vector Machine, Machine Learning, Software Project Management, SHAP Explainability

Zitieren

APA

Radharam, R. (2026). Predicting Task Delays in Distributed Agile Software Development Using Support Vector Machines. International Journal of Agile Software Engineering, 1(1), 1-12. https://doi.org/10.55124/ijas.v1i1.105

Vancouver

Radharam R. Predicting Task Delays in Distributed Agile Software Development Using Support Vector Machines. International Journal of Agile Software Engineering. 2026;1(1):1-12. doi:10.55124/ijas.v1i1.105

BibTeX
@article{Radharam2026,
  author = {Radharam, Rajender},
  title = {Predicting Task Delays in Distributed Agile Software Development Using Support Vector Machines},
  journal = {International Journal of Agile Software Engineering},
  year = {2026},
  volume = {1},
  number = {1},
  pages = {1--12},
  doi = {10.55124/ijas.v1i1.105}
}
Kostenlos starten →