Promoting Agility in Agile: Leveraging Machine Learning for Dynamic User Story Clustering in Scrum

User stories serve to define requirements, but they can pose significant challenges for management, particularly due to their high volume in large-scale software development using agile development methodologies. The problem is further compounded by the possibility of engineers swapping a lot of information directly, as well as the best practice of not duplicating effort in software production. Therefore, we present an approach that applies NLP and sophisticated clustering algorithms to enhance the management of user stories. We utilized Natural Language Processing to identify requirements from a dataset of user stories, employing clusters such as K-means, K-medoids, hierarchical, and distribution-based clusters. We employed silhouette analysis, a popular tool today, to evaluate the quality of the clusters we obtained. It was helpful in providing an overall view of the system’s requirements, as well as in classifying new requirements into different clusters. This work aims to apply this new methodology to certain user stories within mobile enterprise application projects, and it has demonstrated significant potential. This work significantly reduced the time required for software design, resulting in increased productivity through the use of machine learning technologies in the classification of user stories. This work also demonstrates how to incorporate machine learning into user story management for large organizations, thereby reducing handwork and enhancing validity.