Cost-Sensitive Radial Basis Function Neural Network Classifier for Software Defect Prediction

Table 5

Performance measures.

Performance measures	Definition of the measure	Description
Sensitivity (or); Recall (or); True positive rate (or); Probability of detection (pd)		Proportion of defective modules correctly predicted
Precision		Proportion of modules predicted as defective
False positive rate (or); Probability of false alarm (pf)		Proportion of nondefective modules predicted as defective
Specificity		Proportion of correctly predicted nondefective modules
Classification accuracy		Proportion of correctly predicted modules
Balance		Balance combines pf and pd into one measure and is defined as the distance from the ROC “sweet spot” (where pd = 1 and pf = 0)
Receiver Operating Characteristics (ROC) curve	A graphical plot of “pd” versus “pf” where the discrimination threshold is varied