Local versus Global Models for Just-In-Time Software Defect Prediction

Table 7

Local vs. global models in the timewise-cross-validation scenario.

Indicators	Project	Global	Local (k = 2)
AUC	BUG	0.672 ± 0.125	0.545 ± 0.124
	COL	0.717 ± 0.075	0.636 ± 0.135
	JDT	0.708 ± 0.043	0.645 ± 0.073
	MOZ	0.749 ± 0.034	0.648 ± 0.128
	PLA	0.709 ± 0.054	0.642 ± 0.076
	POS	0.743 ± 0.072	0.702 ± 0.095
	W/D/L	—	0/0/6
F1	BUG	0.587 ± 0.108	0.515 ± 0.128
	COL	0.651 ± 0.066	0.599 ± 0.125
	JDT	0.722 ± 0.048	0.623 ± 0.161
	MOZ	0.804 ± 0.050	0.696 ± 0.190
	PLA	0.702 ± 0.053	0.618 ± 0.162
	POS	0.714 ± 0.065	0.657 ± 0.128
	W/D/L	—	0/0/6
ACC	BUG	0.357 ± 0.212	0.242 ± 0.210
	COL	0.426 ± 0.168	0.321 ± 0.186
	JDT	0.356 ± 0.148	0.329 ± 0.189
	MOZ	0.218 ± 0.132	0.242 ± 0.125
	PLA	0.358 ± 0.181	0.371 ± 0.196
	POS	0.345 ± 0.159	0.306 ± 0.205
	W/D/L	—	0/4/2
	BUG	0.622 ± 0.194	0.458 ± 0.189
	COL	0.669 ± 0.150	0.553 ± 0.204
	JDT	0.654 ± 0.101	0.624 ± 0.142
	MOZ	0.537 ± 0.110	0.537 ± 0.122
	PLA	0.631 ± 0.115	0.645 ± 0.151
	POS	0.615 ± 0.131	0.583 ± 0.167
	W/D/L	—	0/4/2