Incorporating Machine Learning into Agile Scrum for Predictive Defect Management

Predictive defect management using ML in Agile Scrum processes shows great promise by spotting possible problems before they affect the final product. Machine learning (ML) in Agile Scrum for predictive defect management significantly enhances the defect detection process, improving both team performance and sprint outcomes. The proposal begins with data collection, where historical sprint data—such as defect logs, team performance, and sprint completion rates— are gathered and preprocessed to ensure quality for analysis and modelling. The core of the methodology involves training ML models—specifically Random Forest, Neural Networks, and Gradient Boosting—to predict defects based on the prepared data. After training, model diagnostics are conducted to evaluate accuracy, precision, recall, and F1-score. Neural Networks outperformed the other models in defect prediction, achieving the highest accuracy (94%), precision (90%), recall (94.7%), and F1-score (92.3%). The integration of ML into Agile Scrum resulted in a significant improvement in defect detection rate, which increased from 62.1% to 80.4%, and reduced defect resolution time from 3.32 days to 2.56 days. The overall defect trend also dropped from 13.6% to 8.2%. Also provide the comparative analysis between existing models decision tree and naïve bayes that perform bad in compare to proposed modes. These results indicate that the use of ML in Agile Scrum not only improves defect management but also optimises sprint outcomes by providing a more efficient, data-driven approach to handling defects.