A State of the Art Regressor Model’s comparison for Effort Estimation of Agile software

Advances and innovations in the field of software engineering are increasing rapidly. This sensitizes researchers to explore the various cross-cutting concerns incorporated to handle the complexities of various domains of interest. One such thrust area is effort estimation in Agile-inspired software. Estimation has always been challenging in an Agile environment because of its requirement volatility. This paper introduces a critical review of state-of-the-art regression techniques to estimate the efforts of Agile projects. It can be concluded from the obtained results that ensemble estimation techniques outperformed single techniques of estimation. The data have been taken from various companies implementing Agile practices. Different regressors have been trained, tested, crossvalidated, and optimized to fill the actual and estimated effort gap. We have used six regression techniques in this paper, Extreme Gradient Boosting (XGB), Decision Tree (DT), Linear Regressor (LR), Random Forest (RF), Adaptive Boosting (AdaBoost) and, Categorical boosting (CatBoost) regressors. Cat Boost regressor wins with the lowest Root Mean Square Error (RMSE) in comparison to other regressors.