| Computer software effort estimation is receiving increased attention due to the rapid growth of software in major information technology development projects. The annual expenditure projection for information technology software is expected to be {dollar}300 billion in the year 2001. Current parametric models for software effort estimation rely heavily upon sparse historical databases, which require calibration based upon each organization's characteristics, which are reflected in factors such as personnel experience, product characteristics, and process maturity. Each of these models relies on the assumption that development effort is predominantly independent of programming language, or at best, related to the development environment in terms of reliance upon modern programming practices.; This dissertation investigates the problem of software effort estimation and the need for programming language specific effort estimation models. Statistically based unique programming language models for Ada, C, FORTRAN, and Pascal are developed utilizing historical data archived at the U.S. Army Aviation and Missile Command's Life Cycle Software Engineering Center. An amalgamated model is developed based upon all four programming languages.; The results of this research indicate that statistically significant models for each of the programming languages are achievable, with an adjusted multiple coefficient of determination (R¯2) value greater than 95%. Unfortunately, it was not possible to develop a statistically significant amalgamated model based upon the requirement R¯ 2 ≥ 80%. This research suggests that programming language specific effort estimation models are highly desirable and may explain why previous effort estimation models frequently underestimated software build opportunities. It is also significant to note that this research acquired statistically significant effort prediction models based upon a small number of independent variables. The developed models were implemented in Visual Basic to expedite the software effort estimation. |
