Abstract—Since most of the projects encounter effort overruns, effort estimation is one of the most important estimates of software projects during software development. There are several software effort estimation methodologies in the literature. However, instead of proposing a novel effort estimation methodology finding the necessary attributes that affects the software effort estimation is an also important contribution. This study focuses on analyzing the necessity of these attributes. We apply linear regression technique to investigate relation between these attributes. Lastly we evaluate our prediction performance with using Magnitude of Relative Error (MRE), Mean Magnitude of Relative Error (MMRE), Median Magnitude of Relative Error (MdMRE), MSE (Mean Square Error) and Prediction Quality (pred(e)). In order to conduct case study we used the Desharnais (77 projects) dataset from the publicly available PROMISE software engineering repository. Results show that the attribute “PointsNonAdjust” is the most necessary attribute in order to estimate software effort.
Index Terms—Desharnais dataset, linear regression, software effort estimation.
T. E. A. and H. C. T. are with the Computer Engineering Department, Başkent University, Ankara, Turkey (e-mail: ercelebi@ baskent.edu.tr, hasancanterzi@ gmail.com).
[PDF]
Cite:Tülin Erçelebi Ayyıldız and Hasan Can Terzi, "Case Study on Software Effort Estimation," International Journal of Information and Electronics Engineering vol. 7, no. 3, pp. 103-107, 2017.
