User Story Clustering using K-Means Algorithm in Agile Requirement Engineering
Bhawnesh Kumar, Umesh Kumar Tiwari, Dinesh C. Dobhal, Harendra Singh Negi
Abstract
In Agile software development user story plays an important role. User story is the concept of requirement engineering that manages requirements in the form of text as software requirement specification maintained in traditional development models. Proper format is used to write the user story which is easily understandable. The impact of poor user story design may lead to modifications in later phases of development. It also places negative impression on the customer, project and development team as well. Therefore the analysis of these user stories is necessary before they are assigned to the development team. In this paper we propose an approach of user story clustering using k-means algorithm to analyze these user stories. Through this approach similar user stories are clustered on the basis of similarity measures. Our experimental results show that as the value of k increases the quality of resulting cluster improves. To generate balanced clusters our approach uses tf-idf which is comparatively better than count vectorizer. This approach reduces the implementation time of requirements and hence the overall development life cycle of the software. To illustrate our approach we use a case study considering various user stories.