Analysis and extension of COCOMO software cost estimation model

In this thesis, selected cost estimation models are considered and their sensitivity to imprecision of their parameter estimates is studied. The results of sensitivity analysis are then used as a starting point to enhance these models to acknowledge and directly manipulate such imprecision. The methods employed here stem from the technology of fuzzy systems, allowing direct expression of the imprecision and manipulating it in a way that provides for enhanced information about the imprecision/usefulness of the resulting estimates.; The first model, developed using fuzzy regression, is interesting but not practical due to very large spread of the estimates. The second model, developed using fuzzy arithmetic, provides cost estimates in the form of fuzzy numbers carrying both the modal value of the estimated effort and a measure of its precision. In this case, the spread of the estimates is reasonable and provides valuable information regarding reliability of the estimates.