An evaluation of effort estimation supported by change impact analysis in agile software development

摘要

In agile software development, functionality is added to the system in an incremental and iterative manner. Practitioners often rely on expert judgment to estimate the effort in this context. However, the impact of a change on the existing system can provide objective information to practitioners to arrive at an informed estimate. In this regard, we have developed a hybrid method, that utilizes change impact analysis information for improving effort estimation. We also developed an estimation model based on gradient boosted trees (GBT). In this study, we evaluate the performance and usefulness of our hybrid method with tool support and the GBT model in a live iteration at Insiders Technologies GmbH, a German software company. Additionally, the solution was also assessed for perceived usefulness and understandability in a study with graduate and post-graduate students. The results from the industrial evaluation show that the proposed method produces more accurate estimates than only expert-based or only model-based estimates. Furthermore, both students and practitioners perceived the usefulness and understandability of the method positively.