In-Depth Analysis and Prediction of Coupling Metrics of Open Source Software Projects

Munish Saini, Raghuvar Arora, Sulaimon Adebayo

OAI: oai:igi-global.com:301267 • DOI: 10.4018/JITR.301267

This research was conducted to perform an in-depth analysis of the coupling metrics of 10 Open Source Software (OSS) projects obtained from the Comets dataset. More precisely, we analyze the dataset of object-oriented OSS projects (having 17 code related metrics such as coupling, complexity, and size metrics) to (1) examine the relationships among the coupling and other metrics (size, complexity), (2) analyze the pattern in the growth of software metrics, and (3) propose a model for prediction of coupling. To generalize the model of coupling prediction, we have applied different machine learning algorithms and validated their performance on similar datasets. The results indicated that the Random forests algorithm outperforms all other models. The relation analysis specifies the existence of strong positive relationships between the coupling, size, and complexity metrics while the pattern analysis pinpointed the increasing growth trend for coupling. The obtained outcomes will help the developers, project managers, and stakeholders in better understating the state of software health