Smart Agile Prioritization and Clustering: An AI-Driven Approach for Requirements Prioritization

In Agile software development, requirements prioritization plays a crucial role in ensuring that critical functionalities are delivered efficiently. Traditional prioritization methods often suffer from scalability limitations, lack of automation, and difficulty in handling dependencies. This paper proposes Smart Agile Prioritization and Clustering (SAPC), an AI-driven approach that enhances requirements prioritization by leveraging Natural Language Processing (NLP), BERT embeddings, graph-based dependency modeling, and optimization techniques. The proposed model extracts and processes textual requirements, constructs a dependency graph to model interrelations, and applies PageRank to compute requirement importance. Feature fusion and dimensionality reduction using Uniform Manifold Approximation and Projection (UMAP) facilitate clustering, while Particle Swarm Optimization (PSO) determines the optimal number of clusters for efficient backlog prioritization. The effectiveness of SAPC is evaluated using functional requirements extracted from Software Requirement Specifications (SRS), product backlogs, and customer requests, along with a benchmark dataset for validation. Various machine learning algorithms are tested, with KNN and Random Forest demonstrating the highest accuracy and lowest Mean Squared Error (MSE), outperforming traditional prioritization techniques. The results highlight the potential of AI-based methods in automating and optimizing backlog management within Agile methodologies, offering a scalable and data-driven approach to enhanced requirements prioritization.