Achieving accurate clone detection results

Abstract

Many existing clone detection approaches produce substantial amounts of clones that are irrelevant to software engineers. These false positives significantly hinder adoption of clone detection in practice and can lead to inaccurate research conclusions. In this paper, we propose clone coupling as an explicit criterion for the relevance of clones and outline a method for clone detection tailoring. The results of a large industrial case study indicate that it can significantly increase clone detection accuracy.