A Systematic Literature Review of Machine Learning Estimation Approaches in Scrum Projects

It is inevitable for any successful IT industry not to estimate the effort, cost, and duration of their projects. As evident by Standish group chaos manifesto that approx. 43% of the projects are often delivered late and entered crises because of overbudget and less required functions. Improper and inaccurate estimation of software projects leads to a failure, and therefore it must be considered in true letter and spirit. When Agile principle-based process models (e.g., Scrum) came into the market, a significant change can be seen. This change in culture proves to be a boon for strengthening the collaboration between developer and customer. Estimation has always been challenging in Agile as requirements are volatile. This encourages researchers to work on effort estimation. There are many reasons for the gap between estimated and actual effort, viz., project, people, and resistance factors, wrong use of cost drivers, ignorance of regression testing effort, understandability of user story size and its associated complexity, etc. This paper reviewed the work of numerous authors and potential researchers working on bridging the gap of actual and estimated effort. Through intensive and literature review, it can be inferred that machine learning models clearly outperformed non-machine learning and traditional techniques of estimation.