Software Project Eeort Estimation Using Genetic Programming Ii Grammar Guided Genetic Programming

Knowing the estimated cost of a software project early in the development cycle is a valuable asset for management. In this paper, an evolutionary computation method, Grammar Guided Genetic Programming (GGGP), is used to t models, with the aim of improving the prediction of software development costs. Valuable results are obtained, signiicantly better than those obtained by simple linear regression. In this research, GGGP, because of its exibility and the ability of incorporating background knowledge, also shows great potential in being applied in other software engineering modeling problems. Knowing the estimated cost of a particular software project early in the development cycle is a valuable asset. Management can use cost estimates to evaluate a project proposal or to manage the development process more eeectively. Therefore, the accurate prediction of software development cost may have a large economic impact: in fact, some 60% of large projects signiicantly overrun their estimates and 15% of the software projects are never completed due to the gross misestimation of development 1].The main driver of cost is eeort. Thus cost estimation is largely a problem of eeort estimation. A large range of metrics have been proposed for early estimation of software project eeort. A number of authors have suggested that the standard sets have too many parameters, and a number of reduced sets have been suggested (large metric sets have high collection costs, and also risk generating over-tted models). The reductions have relied on linear methods to eliminate metrics, and linear models for estimating size and eeort from the metric sets, but there is a risk that some of the dependencies may be non-linear. Researchers elsewhere have begun to investigate alternative methods of developing predictive models, including fuzzy logic, Corresponding author neural networks, and regression trees. Exploration of evolutionary approaches has just begun. 2, 3]. In this paper, an evolutionary approach, Grammar Guided Genetic Programming (GGGP), is used to t nonlinear models to a dataset of past projects, aiming to determine appropriate metric sets and improve the prediction of software development eeort. In the following Section 2, GGGP is introduced very brieey. The application of GGGP in evolution of software development eeort estimation programs is discussed in Section 3. This includes data preparation, GP details and results obtained. The results are analyzed in Section 4. Section 5 draws conclusions. One limitation of canonical Genetic Programming (GP) 4, 5] is its requirement of closure. It implies that …