An empirical study on clone consistency prediction based on machine learning

摘要

Context: Code Clones have been accepted as a common phenomenon in software, thanks to the increasing demand for rapid production of software. The existence of code clones is recognized by developers in the form of clone group, which includes several pieces of clone fragments that are similar to one another. A change in one of these clone fragments may indicate necessary & ldquo;consistent changes & rdquo;are required for the rest of the clones within the same group, which can increase extra maintenance costs. A failure in making such consistent change when it is necessary is commonly known as a & ldquo;clone consistency-defect & rdquo;, which can adversely impact software maintainability. Objective: Predicting the need for & ldquo;clone consistent changes & rdquo;after successful clone-creating or clone-changing operations can help developers maintain clone changes effectively, avoid consistency-defects and reduce maintenance cost.Method: In this work, we use several sets of attributes in two scenarios of clone operations (clone-creating and clone-changing), and conduct an empirical study on five different machine-learning methods to assess each of their clone consistency predictability & mdash; whether any one of the clone operations will require or be free of clone consistency maintenance in future.Results: We perform our experiments on eight open-source projects. Our study shows that such predictions can be reasonably effective both for clone-creating and changing operating instances. We also investigate the use of five different machine-learning methods for predictions and show that our selected features are effective in predicting the needs of consistency-maintenance across all selected machine-learning methods.Conclusion: The empirical study conducted here demonstrates that the models developed by different machine-learning methods with the specified sets of attributes have the ability to perform clone-consistency prediction.