software module. After this phase, this model is used to predict the faultproneness
labels of modules that locate in a new software version. Figure 1 shows
this fault prediction process. Recent advances in software fault prediction allow
building defect predictors with a mean probability of detection of 71 percent and
mean false alarm rates of 25 percent [29]. These rates are at an acceptable level
and this quality assurance activity is expected to quickly achieve widespread
applicability in the software industry.
Figure