Optimal Additive C-Fuzzy Regression Trees for Software Development Effort Prediction

Abstract

There is evidence that precise prediction of the software development effort plays a crucial role in properly monitoring and managing software projects. Researchers have suggested many software effort estimation techniques. Nonetheless, none of these methods performed well in all circumstances. Ensemble models have been recently proposed in the literature to overcome the significant drawbacks of single machine learning approaches. In this study, we proposed a novel model, the ensemble of optimal additive cluster-based fuzzy regression trees for software development effort prediction. We performed an empirical evaluation using four datasets and the 30% holdout cross-validation technique. We compared the performance of our proposed ensemble model to the c-fuzzy regression tree, the bagged c-fuzzy regression tree model, the ensemble of optimal trees, random forest, and regression trees. Our suggested model outperforms all the compared models in Pred (25%), MMRE, and MdMRE in all employed datasets.