The widespread use of software metrics to predict various quality attributes is evident. In this study we have used metrics to identify change prone parts of software so that developers can pay focused attention on such classes. Various metric models are constructed using machine learning and statistical techniques, which can be used for predicting change prone parts of the software. However, training these models is a time consuming task and hence, these models cannot be used on a daily basis to predict change proneness. In this paper, an alternative approach is used which is based on calculating thresholds of metrics. Thresholds are defined as alarming values above which a class is considered to be risky or change prone and hence, needs careful attention. A statistical approach is used to calculate threshold values of open source software, Freemind 0.9.0. To examine the applicability of threshold values, they are validated on the different releases of Freemind as well as on a similar nature project, Frinika. The results demonstrate the effectiveness of the methodology in identification of the threshold values.
展开▼
