Cross-Project Defect Prediction Based on Two-Phase Feature Importance Amplification

Table 2

F1-measure, AUC, and MCC of model with filtering (Model 1) and model without filtering (Model 2).

	F1-measure		AUC		MCC
	Model 1	Model 2	Model 1	Model 2	Model 1	Model 2
EQ-JDT	0.415	0.790	0.632	0.873	0.232	0.735
EQ-LC	0.300	0.761	0.726	0.903	0.265	0.738
EQ-ML	0.242	0.742	0.545	0.875	0.061	0.702
EQ-PDE	0.291	0.776	0.600	0.864	0.146	0.740
JDT-EQ	0.276	0.833	0.576	0.862	0.291	0.719
JDT-LC	0.322	0.752	0.602	0.875	0.330	0.726
JDT-ML	0.283	0.745	0.584	0.883	0.229	0.707
JDT-PDE	0.247	0.761	0.569	0.866	0.233	0.721
LC-EQ	0.196	0.831	0.554	0.862	0.261	0.710
LC-JDT	0.592	0.816	0.727	0.911	0.516	0.768
LC-ML	0.347	0.765	0.626	0.855	0.243	0.732
LC-PDE	0.256	0.787	0.572	0.866	0.234	0.755
ML-EQ	0.183	0.855	0.550	0.886	0.251	0.755
ML-JDT	0.396	0.787	0.623	0.883	0.397	0.730
ML-LC	0.219	0.824	0.562	0.925	0.316	0.806
ML-PDE	0.217	0.763	0.559	0.865	0.229	0.724
PDE-EQ	0.238	0.873	0.560	0.897	0.232	0.785
PDE-JDT	0.477	0.806	0.666	0.881	0.366	0.754
PDE-LC	0.372	0.784	0.620	0.873	0.397	0.763
PDE-ML	0.292	0.754	0.594	0.853	0.178	0.718
Average	0.308	0.790	0.602	0.878	0.270	0.739

The values in bold are results with the best performance of each instance.