Cluster Analysis in the Identification of Patterns in Software Development with Agile Methodologies: A Systematic Literature Review

Software Engineering is a discipline that aims to deliver quality software products. To this end, several development methodologies have been proposed, with agile methodologies being the most recently used, which promote the continuous delivery of functional software in short cycles called iterations or sprints. In these deliveries, there are activities where the identification of patterns in different software arti-facts is necessary. An example of this is the detection of similar artifacts to avoid redundancy or reuse artifacts. This study presents a systematic review of the literature that seeks to analyze the activities, development stages, benefits, and challenges that have been reported when applying various cluster analysis algorithms to identify patterns in agile methodologies. Fifteen primary studies selected through the Kitchenham methodology were analyzed. It was observed that k-means, k-medoids, and hierarchical clustering are the most used approaches in various stages of development such as requirements, design, and testing. Furthermore, the most analyzed artifacts were user stories. Finally, the studies reveal various advantages, such as pattern identification without a priori knowledge, variance reduction, and development time reduction, among others. However, the parameterization of these algorithms is still a fundamental challenge to be addressed in future work.