A Hybrid Technique using Grey Relational Analysis and Regression for Software Effort Estimation using Feature Selection
Geeta Nagpal1, Moin Uddin2, Arvinder Kaur3
1Geeta Nagpal, Lecturer, Department of CS, National Institute of Technology, Jalandhar, India.
2Moin Uddin, Pro Vice Chancellor, Delhi Technological University, Delhi, India.
3Arvinder Kaur, Associate Professor, University School of IT, GGSIPU, Delhi, India.
Manuscript received on November 19, 2011. | Revised Manuscript received on November 29, 2011. | Manuscript published on January 05, 2012. | PP: 20-27 | Volume-1 Issue-6, January 2012. | Retrieval Number: F0256101511/2012©BEIESP
Open Access | Ethics and Policies | Cite
© The Authors. Published By: Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Abstract: Software Estimation Techniques present an inclusive set of directives for software project developers, project managers and the management in order to produce more accurate estimates or predictions for future developments. The estimates also facilitate allocation of resources’ for Software development. Estimations also smooth the process of re-planning, prioritizing, classification and reuse of the projects. Various estimation models are widely being used in the Industry as well for research purposes. Several comparative studies have been executed on them, but choosing the best technique is quite intricate. Estimation by Analogy(EbA) is the method of making estimations based on the outcome from k most analogous projects. The projects close in distance are potentially similar to the reference project from the repository of projects. This method has widely been accepted and is quite popular as it impersonates human beings inherent judgment skill by estimating with analogous projects. In this paper, Grey Relational Analysis(GRA) is used as the method for feature selection and also for locating the closest analogous projects to the reference project from the set of projects. The closest k projects are then used to build regression models. Regression techniques like Multiple Linear Regression, Stepwise Regression and Robust regression techniques are used to find the effort from the closest projects.
Keywords: Estimation by Analogy, Feature Selection, Grey relational Analysis, Regression.