Abstract
At an early stage in the development process, a development team must obtain insight into the software being developed to establish a reliable plan. Thus, the team members should investigate, in depth, any information relating to the development. A major challenge for developers is software development effort estimation (SDEE), which refers to gauging the amount of effort needed to develop the software. In agile methodologies, a project is delivered in iterations, each of which delivers a set of requirements known as user stories. Therefore, SDEE in agile focuses on estimating a single user story’s effort, not the project as a whole, as in traditional development. Among the various techniques, data-driven methods have proved effective in effort estimation, as they are unaffected by external pressure from managers. Moreover, no experts have to be available at the point when estimation is undertaken. By conducting a systematic literature review, this study presents a comprehensive overview of data-driven techniques for user story effort estimation. The results show that there has been limited work on this topic. Studies were analysed to address questions covering five main points: technique; performance evaluation method; accuracy, independent factors (effort drivers); and the characteristics of the datasets. The main performance evaluation methods are performance measures, baseline benchmarks, statistical tests, distribution of estimates, comparison against similar existing techniques and human estimation. Four types of independent factors were identified: personnel; product; process; and estimation. Furthermore, the story point was found to be the most frequently used effort metric in agile user stories.