Prediction of Software Effort Using Design Metrics: An Empirical Investigation

Abstract

These days, prediction of effort in software project is the shove area for the researchers. The estimation of effort in software process is as essential as software product. Primarily, estimation models consist of relation between dependant and independent variable(s). The effectiveness of these models is to bring more accuracy to the work plan and reduce financial cost. The variables in these models may be considered as complexity, size, person per month, and other different software metrics. Most of these models only considered the static behaviour of the software product, in which the fixed value of the effort predicted at the starting of project. Hence, there is a need to formulate a methodology which considered the future changes in the software project for effort estimation. In this paper, a model has been formulated which can be use to make the prediction of software efforts with the help of software metrics, primarily design metrics, such as Depth of Inheritance Tree, Line of Code, Weighted Method per Class. The correlation between the metrics and effort is been shown with the help of regression model formulated in this paper. The model has been validated by the data set collected from the PROMISE repository.