A Software Defect Prediction Method Based on Program Semantic Feature Mining

As the size and complexity of software systems grow, knowing how to effectively judge whether there are defects in programs has attracted extensive attention research. However, current defect prediction methods only extract semantic information at syntactic level lack features mine manifestations code, because defective is incomplete or representation. Defective exhibits flawed behavior. This paper proposes a method based on program semantics feature mining (PSFM) method. Specifically, first extracted from code grammatical structure text information. Then, mined through Finally, predicted by using features. The experimental results show that, compared with existing methods, this (PSFM method) obtained higher F-measure value.