A Software Defect Prediction Approach Based on Hybrid Feature Dimensionality Reduction

Table 1

Experimental results of two kinds of dimensionality reduction methods.

Projects	Metrics	ALL	IG	KPCA
Ant	AUC	0.924	0.924	0.944
	F1	0.939	0.939	0.960
	MCC	0.854	0.854	0.906
Arc	AUC	0.772	0.772	0.736
	F1	0.782	0.787	0.756
	MCC	0.547	0.551	0.480
Camel	AUC	0.964	0.978	0.957
	F1	0.961	0.976	0.954
	MCC	0.927	0.955	0.912
Poi	AUC	0.776	0.790	0.790
	F1	0.720	0.735	0.735
	MCC	0.532	0.559	0.559
Redktor	AUC	0.801	0.777	0.869
	F1	0.778	0.745	0.852
	MCC	0.598	0.557	0.733
Tomcat	AUC	0.850	0.853	0.850
	F1	0.841	0.853	0.850
	MCC	0.699	0.714	0.708
Velocity	AUC	0.782	0.780	0.780
	F1	0.787	0.794	0.794
	MCC	0.563	0.558	0.558
Xalan	AUC	0.807	0.817	0.838
	F1	0.789	0.804	0.828
	MCC	0.615	0.633	0.675
Xerces	AUC	0.789	0.544	0.795
	F1	0.795	0.788	0.792
	MCC	0.611	0.601	0.596
AVE	AUC	0.829	0.830	0.840
	F1	0.821	0.824	0.836
	MCC	0.661	0.665	0.681