Symmetric bug prediction in software requirement by machine learning algorithms - Scientific Reports

▤ Summary

This article examines how machine learning can predict software defects and estimate bug resolution time. It compares Random Forest, ANN, and KNN models using JIRA and NASA datasets, then uses feature selection and PCA to reduce noise and improve accuracy. The study highlights that structured data from the NASA dataset produces stronger predictive performance than the more random JIRA dataset. It also shows how the selected features, such as State Change and Execution Time Hour, help explain defect inflow and resolution behavior. The work is relevant to software quality and development workflow optimization, but it is presented as research rather than a commercial product update.

▦ Linked Companies