Early software size estimation is a challenging task since limited information is available at the time of project inception. Additional information, however, is gradually added as development progresses. The goal of this research is to quantitatively capture the impact on early software size estimation of this additional information introduced especially when transitioning from the analysis phase to the design phase by comparing the analysis class diagram (ACD) and the design class diagram (DCD). We introduce a new class of metrics called analysis-to-design adjustment factors (ADAFs) to accomplish this goal. ADAFs are calculated for four different class diagram metricsnumber of classes (NOC), number of attributes (NOA), number of methods (NOM), and number of relationships (NOR)used in different class diagram-based software size estimation models. We use practical, theoretical, and empirical validation methods to evaluate the applicability of these ADAFs. To assess the utility of these ADAFs in early software size estimation, we compare the accuracy of existing early software size estimation models before and after the application of ADAFs. Results indicate a marked improvement in the accuracy of these models after the application of ADAFs. Furthermore, regression-based models employing problem domain metrics have also been built to predict these ADAFs. All of these models are statistically significant (pvalues < 0.05) with R 2 values between 0.42 and 0.88.

No abstract

No abstract

Software size estimation is a vital activity of software project planning and management. Early software size estimation is a challenging task due to the limited information available during the early phases of software development. The goal of this paper is to construct and validate early software size estimation models based on four analysis-to-design adjustment factor (ADAF)-adjusted analysis class diagram metrics (i.e. ADAF-adjusted number of classes, ADAF-adjusted number of attributes, ADAF-adjusted number of methods and ADAF-adjusted number of relationships) using stepwise multiple linear regression and leave-one-out cross-validation. Furthermore, the prediction accuracy of the best-performing proposed model is also compared with the model based on objective class points. The results of this comparison reveal that our proposed method reduces errors significantly (i.e. on average, 16% reduction in mean absolute residual and 24% reduction in mean squared error).