Correlations Between Problem Domain and Solution Domain Size Measures for Open Source Software

Abstract

Predicting how much effort will be required to complete a software project as early as possible is a very important factor in the success of software development projects. Including function points and its variants, there are several size measures and corresponding measurement methods that can be used for effort estimation. However, in most of the projects, there is limited amount of information available in the early stages and significant effort is spent for size measurement and effort estimation with such methods. This paper analyzes the correlation between the size metrics of conceptual model of the problem domain and the resulting software. For this purpose, we consider open source project management and game software. We apply linear regression and cross validation techniques to investigate the relation between the sizes of problem domain (i.e., conceptual) and solution domain (i.e., design) models. The results reveal a high correlation between the number of conceptual classes in the problem domain model and the number of software classes constituting the corresponding software. The results suggest that it is possible to use problem domain descriptions in the early stages of software development projects to make plausible predictions for the size of the software.