An Empirical Study of Software Metrics Diversity for Cross-Project Defect Prediction

Table 6

Comparison between the proposed approach and two baseline methods: TCA+ and TDSelector.

Source/target	TCA+	TDSelector	CPDP-OSS
Eclipse/Equinox	0.600 (+5.9%)	0.514 (+23.7%)	0.636
Lucene/Equinox	0.620 (+4.4%)	0.664 (−2.5%)	0.648
Mylyn/Equinox	0.560 (−11.6%)	0.459 (+7.8%)	0.495
Pde/Equinox	0.600 (+4.0%)	0.610 (+2.3%)	0.624
Equinox/Eclipse	0.540 (−4.5%)	0.519 (−0.7%)	0.516
Lucene/Eclipse	0.560 (−4.5%)	0.594 (−9.9%)	0.535
Pde/Eclipse	0.430 (+27.7%)	0.455 (+20.7%)	0.549
Mylyn/Eclipse	0.480 (+13.5%)	0.510 (+5.9%)	0.545
Equinox/Lucene	0.270 (+50.6%)	0.337 (+20.7%)	0.407
Eclipse/Lucene	0.310 (+26.9%)	0.313 (+25.7%)	0.393
Mylyn/Lucene	0.250 (+32.0%)	0.298 (+10.7%)	0.330
Pde/Lucene	0.330 (−4.9%)	0.314 (+0.0%)	0.314
Equinox/Mylyn	0.230 (+24.7%)	0.264 (+8.6%)	0.287
Eclipse/Mylyn	0.360 (−13.1%)	0.303 (+3.2%)	0.313
Lucene/Mylyn	0.290 (+6.4%)	0.291 (+6.1%)	0.309
Pde/Mylyn	0.290 (−6.1%)	0.300 (−9.2%)	0.272
Equinox/Pde	0.330 (+21.2%)	0.367 (+9.0%)	0.400
Eclipse/Pde	0.380 (+2.8%)	0.391 (+0.0%)	0.391
Lucene/Pde	0.370 (−11.8%)	0.365 (−10.6%)	0.326
Mylyn/Pde	0.370 (+0.4%)	0.371 (+0.0%)	0.371
Average	0.410 (+6.0%)	0.412 (+5.1%)	0.433