K-Means and Feature Selection Mechanism to Improve Performance of Clustering User Stories in Agile Development

Agile development is an approach in software development that aims to improve the flexibility, responsiveness, and quality of software products. One part of the planning phase in agile development is writing user stories. User story is a text used to detail needs from the user's point of view. The semi-natural and simplicity of user stories sometimes results ambiguity, inconsistent sentences, and incomplete sentences. This ambiguous user story has the potential to cause problems or conflicts, especially when programmers start building the product. To solve that problem, this research. To solve this problem, a method that can identify ambiguous and problematic user stories is needed. One method that can be implemented to solve this problem is the clustering method. This research uses the K-means algorithm because K-means has proven to be very good at clustering data. To improve clustering performance, we have added Variance Threshold method to select features. The Silhouette Value evaluation results in this experiment also show a significant increase to reach 0.9634384171608066 in clustering experiments with feature selection. This means that the feature selection method does have a positive impact on the clustering evaluation results.