Software cost estimation is an important activity during software development. There are many sophisticated parametric models for estimating the size, cost, and schedule of Object Oriented (OO) software projects. Several authors assert that a model's predictive accuracy can be improved by calibrating its default parameters to a specific environment.. The main aim of this paper is to improve the cost estimation accuracy of object oriented console applications at design phase. This paper suggests a variation on the existing formula by calibrating COCOMOII early design model for a specific organization environment using Object Oriented Function Point (OOFP) as a size measure instead of standard Function Point (FP) that is used currently in COCOMOII model. This approach benefits from the model calibration and considers OO aspects such as inheritance, aggregation, association and polymorphism. Based on historical projects of an organization, linear regression is applied on the software projects data to relate LOC to software OOFP. The obtained regression model is then used to estimate LOC as a step towards estimating effort and cost for new projects. The authors developed a software tool to implement the proposed estimation approach. Experiment with the cost estimation tool showed that it presents a user friendly automatic cost estimation for OO console applications and generates reasonable results.
Key words
Object Oriented, Software Cost Estimation, COCOMO, Calibration, Linear Regression.
|
