Using Object-Oriented Design Metrics to Predict Software Defects 1*

Many object-oriented design metrics have been developed [1,3,8,17,24] to help in predict software defects or evaluate design quality. Since a defect prediction model may give crucial clues about the distribution and location of defects and, thereby, test prioritization, accurate prediction can save costs in the testing process. Considerable research has been performed on defect prediction methods; see the surveys by Purao and Vaishnavi [22] and by Wahyudin et al. [25], unfortunately few results appear at statistically significant level. Therefore, further empirical validation is necessary to prove the usefulness of the metrics and software prediction models in industrial practice. Our study was made possible through the creation of a new metric calculation tool. There are many tools that calculate object-oriented metrics. What is the reason to create another one? In fact the situation is not so perfect. The available programs are either extremely inefficient (sometimes they do not work with big software projects at all), not available as open source and therefore difficult to reason about their results, or incomplete — the set of calculated metrics is not wide enough. It is extremely hard to find a tool that calculates all metrics from the Chidamber and Kemerer (C&K) metrics suite [3]. Having both, C&K and QMOOD metrics suites [1] in one tool is even rarer, and according to the authors' knowledge there is no other tool, that calculates metrics suggested by Tang et al. [24]. Ckjm calculates metrics that have been recommended as good quality indicators. There are several works that investigate the C&K metric suite and that have empirically proven their usability in quality or defect prediction [2, 10, 11, 20]. There are recommendations about QMOOD metrics suite [1, 20] and the quality oriented extension of C&K [24] too. Ckjm does not offer a GUI and its focus is not on