Identifying Agile Roles in Software Engineering Projects using Repository and Work-Tracking Data

Software development at corporate scale is a collaborative effort requiring the involvement of people with different responsibilities. In agile software engineering approaches like SCRUM, responsibilities are divided along two main axes. Team membership sets the scope of a group of engineers, defining the parts of the software or types of challenges the individual approaches. Roles define the contribution of the individual within the team. As team and role data are not necessarily defined in the analyzed repository and work-tracking systems, their prediction from available data is beneficial for the understanding of processes and the providing of stakeholder-focused insights based on data. In this paper, we show that data gathered from software repositories and work-tracking systems can be used to predict the role of persons within these systems using machine learning, achieving an accuracy of around 83 percent. We also show how the process definition, specifically the work item hierarchy of the analyzed system is reflected in the analyzed data.