Abstract
Software effort prediction is an important and challenging activity that takes place during the early stages of software development, where costing is needed. Software size estimate is one of the most popular inputs for software effort prediction models. Accordingly, providing a size estimate with good accuracy early in the lifecycle is very important; it is equally challenging too. Estimates that are computed early in the development lifecycle, when it is needed the most, are typically associated with uncertainty. However, none of the prominent software effort prediction techniques or software size metrics addresses this issue satisfactorily. In this paper, we propose a framework for developing probabilistic size proxies for software effort prediction using information from conceptual UML models created early in the software development lifecycle. The framework accounts for uncertainty in software size and effort prediction by providing the estimate as a probability density function instead of a certain value. We conducted a case study using open source datasets and the results were encouraging. (C) 2012 Elsevier B.V. All rights reserved.