User Story Splitting in Agile Software Development using Machine Learning Approach

User story forms the requirements for agile software development in textual format. The user story should be in the proper template so that developers can implement it easily. The quality of the user story should also be maintained semantically. As semantics gives meaningful artifacts about the customer requirements. User stories can be written easily, if there is complexity then need attention to make it easy. As user story splitting techniques are available to split various user stories. In this paper, we proposed a user story-splitting technique in agile software development using a machine learning approach. Data entry and CRUD-based user story-splitting techniques are used to split the complex user story. Also, user story distribution uses some machine learning approaches such as K-medoids and Kmeans clustering based on similarity measures. To measure the performance of the approach, silhouette value is calculated which shows the cohesion between user stories. As result shows that K-means gives the value 0.114 for cluster size 15 whereas, K-medoids gives 0.068 value. K-means is far better than K-medoids clustering. This approach reduces the software development time of requirements and improves performance as well.