Entropy-based framework dealing with error in software development effort estimation

Software engineering community often investigates the error concerning software development effort estimation as a part, and sometimes, as an improvement of an effort estimation technique. The aim of this paper is to propose an approach dealing with both model and attributes measurement error sources whatever the effort estimation technique used. To do that, we explore the concepts of entropy and fuzzy clustering to propose a new framework to cope with both error sources. The proposed framework has been evaluated with the COCOMO'81 dataset and the Fuzzy Analogy effort estimation technique. The results are promising since the actual confidence interval percentages are closer to those proposed by the framework. © 2017 by SCITEPRESS - Science and Technology Publications, Lda. All Rights Reserved.