Bootstrap Confidence Intervals for Regression Error Characteristic Curves Evaluating the Prediction Error of Software Cost Estimation Models

The importance of Software Cost Estimation at the early stages of the development life cycle is clearly portrayed by the utilization of several algorithmic and artificial intelligence models and methods, appeared so far in the literature. Despite the several comparison studies, there seems to be a discrepancy in choosing the best prediction technique between them. Additionally, the large variation of accuracy measures used in the comparison procedure constitutes an inhibitory factor which complicates the decision-making. In this paper, we further extend the utilization of Regression Error Characteristic analysis, a powerful visualization tool with interesting geometrical properties in order to obtain Confidence Intervals for the entire distribution of error functions. As there are certain limitations due to the small-sized and heavily skewed datasets and error functions, we utilize a simulation technique, namely the bootstrap method in order to evaluate the standard error and bias of the accuracy measures, whereas bootstrap confidence intervals are constructed for the Regression Error Characteristic curves. The tool can be applied to any cost estimation situation in order to study the behavior of comparative statistical or artificial intelligence methods and test the significance of difference between models.