Аннотация
Software requirement classification (SRC) is a critical yet challenging task in large-scale software development, where manual classification is time-consuming, error-prone, and unscalable, consuming significant project effort as reported by industry surveys. The urgent need for automated, scalable solutions motivates this research, which proposes a novel integration of advanced machine learning (ML) techniques and a cloud-based architecture to enhance SRC using the PROMISE dataset. Our approach leverages a hybrid cloud–edge deployment strategy, combining the precision of ML models, such as BERT, with dynamic resource allocation to achieve an F1-score of 89.2%, outperforming traditional methods. Key contributions include: (1) a comprehensive evaluation of five ML models for SRC, (2) a novel hybrid cloud–edge architecture balancing performance, latency, and privacy, and (3) a cost–benefit analysis demonstrating cost-effectiveness for enterprise applications. These advancements address scalability and accuracy challenges in requirement engineering, enabling more efficient, consistent, and automated SRC processes, with significant potential for widespread industry adoption. • Cloud-based ML improves scalability of software requirements classification. • Hybrid cloud–edge deployment ensures performance, privacy, and low latency. • Evaluation of 5 ML models on PROMISE dataset with F1-score up to 89.2%. • Proposed solution is cost-effective and industry-ready for SRC automation.