Software fault prediction based on one-class SVM
Lin Chen, Bin Fang, Zhaowei Shang · 2016
Software fault prediction (SFP) is useful for helping the software engineer to locate potential faulty modules in software testing more easily, so that it can save a lot of time and budgets to improve the software quality. In this paper, aiming at solving the problem that the faulty samples are too rare to train a classifier, an one-class SFP model is proposed by using only non-faulty samples based on one-class SVM. The empirical validation is conducted on 6 extremely imbalanced datasets collected from real-world software containing only small amounts of faulty instances. The test results suggest that the proposed model can achieve a reasonable fault prediction performance when using only a small proportion of training samples, and performs much better than conventional and class imbalanced learning based SFP models in terms of G-mean measure. Thus the proposed model provided a considerable solution for SFP with a few faulty modules in early life of software testing.