Abstract
As one of the most important internal attributes of software systems, the estimation of software size is crucial to project management and schedule creation/tracking. Program size can be typically described by the length, functionality, or complexity of the software, but in practice most people customarily use the lines of code (LOC) as a major measure of program size. In actuality, LOC is still widely used, and can be easily measured upon project completion. In this study, we used the methods of multimodel combinations to predict and analyze the size distribution and size-change rate of Open-Source Software (OSS). Experiments based on real OSS data will be performed, and the evaluation results have shown that the proposed method had a better prediction capability of size distribution and size-change rate of OSS. Our goal is not to add one more model to the already-existing large number of models, but to emphasize a new approach for the evaluation of program size variation and reveal different issues of software sizing.