A Novel Effort Estimation Framework for Agile Based Projects

Abstract

Software estimation is the most essential activity in project management. Numerous researchers across the globe have worked on the issue of software effort estimation and have contributed significantly. With advancements in technology and software process models, the old estimation methods may not yield fruitful results for project managers. There is a need to reframe the estimation process in Agile-based project development. We have proposed a novel continuous estimation framework named Agilator to assist project managers involved in software estimation-oriented tasks. This framework provides two novel features. The first one is the auto-adjustment of effort through learning gain accumulated and adjusted from errors deduced during iterations. This feature makes the system end-to-end trainable, laying the foundation for a continuous estimation framework. The second feature is real-time prediction available for Scrum masters. The proposed framework will not replace existing expert-based estimation; instead, it will assist by participating and contributing AI-assisted input to the team. This paper helps to minimize the estimated effort and actual effort for various system stakeholders. We have presented the Agilator framework using ANFIS-EEBAT i.e., Adaptive Neuro-fuzzy Inference System – Energy Efficient BAT algorithm.