Enhancing Requirements Classification Using Machine Learning Techniques

Enhancing Requirements Classification Using Machine Learning Techniques

연구 분야: Artificial Intelligence

학회: SN Computer Science

The software development life cycle (SDLC) is constructed upon software requirements, a reflection of stakeholder needs and expectations. Software quality assurance and guiding future development processes are conditional upon the accurate identification and classification of these requirements into non-functional and functional categories. Although prevalent, manual classification is subjective, time-consuming, and often unreliable. This research explores the application of machine learning to automatically classify software requirements, comparing five algorithms: Random Forest, SVM, KNN, XGBoost, and a Voting Ensemble, on an actual-world dataset labelled with over 13 categories of requirements. The findings show that Random Forest is outstanding, with a classification accuracy of 94.1%. This work presents a scalable and high-accuracy classification pipeline that improves existing practices for automated requirements engineering.