On Software Defect Prediction Using Machine Learning

Table 4

Comparison of AUC between six classifiers. The ranks in the parentheses are used in computation of the Friedman test.

	APLSC	KPCAC	AdaBoost	SMOTE	AKPCAC	AKPLSC
ar3	0.626 (3)	0.588 (5)	0.580 (6)	0.590 (4)	0.699 (1)	0.682 (2)
ar4	0.563 (5)	0.600 (3)	0.555 (6)	0.610 (2)	0.671 (1)	0.579 (4)
ar5	0.626 (5)	0.710 (2)	0.614 (6)	0.651 (3)	0.722 (1)	0.640 (4)
cm1	0.611 (4)	0.724 (1)	0.589 (5)	0.550 (6)	0.681 (2)	0.650 (3)
kc1	0.682 (4)	0.592 (6)	0.627 (5)	0.700 (3)	0.800 (1)	0.768 (2)
kc2	0.591 (6)	0.601 (5)	0.796 (1)	0.635 (3)	0.732 (2)	0.610 (4)
kc3	0.598 (5)	0.569 (6)	0.698 (2)	0.612 (4)	0.658 (3)	0.713 (1)
mw1	0.587 (5)	0.602 (4)	0.534 (6)	0.654 (2)	0.725 (1)	0.639 (3)
pc1	0.692 (6)	0.718 (5)	0.769 (3)	0.753 (4)	0.841 (2)	0.882 (1)
Sum of ranks	𝑇1=43	𝑇2=37	𝑇3=40	𝑇4=31	𝑇5=14	𝑇6=24