GUI-point analysis: a new model for software size estimation
II Oral McNeil Marple · 1998
Software size estimation is the process applied to develop a prediction for the amount of source code required to implement a software application. This estimate is fundamental to determining the resources and schedule needed to implement a system. Present estimation approaches are based upon research and model from the 1970's and do not embrace present day life cycle models nor engineering tools. A new model for software size estimation is introduced within this dissertation to accommodate graphical user interfaces and life cycle approaches utilizing business rules. This GUI-Point model is developed based upon the presentation of data to the user, business rules, and data storage. In order to determine appropriate weights for each of these factors, a sample set of six applications is used. Each application is analyzed by counting the number of lines of source used to implement each GUI object, business rule, and data access. Weighting coefficients are calculated using these counts. The resultant GUI-point contains three factors to accommodate user interfaces, one factor for business rules, and two factors for data storage. This GUI-Point model is particularly appropriate for life cycle models utilizing rapid prototyping techniques that quantify the user interface as part of system requirement analysis. GUI-Point modeling has been calibrated such that it can replace function point estimates in other cost estimation models. In order to validate the model graphs are used to visually confirm the assumption of model linearity and a calculation of error performed. The GUI-Point model is compared to the function point model using a sample set of thirty-two applications. A null hypothesis is then constructed claiming no difference between function point and GUI-point accuracy. To test this claim the coefficient of determination for both methods is calculated. A 95% confidence interval is calculated and the null hypothesis rejected in favor of the alternative hypothesis that the GUI-point model provides a better estimate. In this comparison the GUI-point model is found to provide more accurate prediction of code size.