IFPUG Function Points to COSMIC Function Points convertibility: A fine-grained statistical approach

Abstract

BackgroundFunctional size measurement is widely used in software organizations because it supports the estimation of software development effort. Function Point Analysis was the first functional size measurement method and became quite popular. The COSMIC method is considered a second-generation method, due to its novel design, and has also gained wide acceptance. Since the proposal of the COSMIC method, the measure convertibility issue arose. Many studies have investigated this issue: several conversion techniques have been proposed and their accuracy has been evaluated through empirical studies. ObjectiveThe goal of the paper is to explore statistic conversion criteria that leverage the similarity between the Base Functional Components of the considered functional measurement methods, especially concerning elementary processes and functional processes. MethodStatistical models of the relationship between the considered measures were built, using Least Median of Squares linear regression. The models use measures of Function Point Analysis Base Functional Components and COSMIC Base Functional Components as independent and dependent variables, respectively. Accuracy of conversion was assessed via leave-one-out cross validation. ResultsThe proposed method was tested on three datasets, and was compared with other conversion methods. The proposed method achieved results that are never less accurate – and sometimes much more accurate – than alternative methods’. ConclusionsThe proposed method requires that when traditional Function Points are measured, information concerning Base Functional Components are recorded. If such information is available, the proposed approach is – according to the collected evidence – preferable to other conversion methods, with respect to both the effort required to obtain the results and their accuracy.