In this thesis, a case study on software cost estimation in the development environment of the microprocessor-embedded systems is conducted. Data of nine projects from one organization are collected and analyzed. Regression and correlation analysis are performed to derive the relationships among cost related variables and to show the degree of relation. Multivariate linear regression method is used to analyze the importance of effects of software cost factors.
The number of modules is presented as an alternative measure of product size to the delivered source lines of code. A cost model COCOMO is evaluated by applying this model to the project actuals. A cost estimation tool COST is used for the evaluation. A calibrated version of COCOMO is presented as a reasonable model in a specific organization.