Software Defect Prediction Based on Source Code Stabilization Model
Semyon Igorevich Kirnosenko, Vitaly A. Egunov, Andrey E. Andreev, Dmitriy Nikolaevich Zharikov · 2013
The ability to predict which files in a software system are most likely to contain the largest numbers of faults in the next release can be a very valuable asset for quality increasing before release. There are many special prediction methods for such purpose. Most of them operate with black-box calculation models that do not give interpretation of actions during prediction process. We propose a source code stabilization model. It allows to describe programming process in form of theory of probability task and to estimate probability of presence of logical errors in various source code fragments. We show applicability of this model for building defect prediction model. The results of experiments show that such defect prediction model may be accurate enough for practical usage in industrial environment. We also consider some possible ways of model improvements in context of changes they require.