Balanced Adversarial Tight Matching for Cross-Project Defect Prediction

Table 6

The results of the AUC comparison with nine methods.

Target project	LR	NNFilter	TCA	TCA+	DBN	DPCNN	TCNN	MANN	ADA	Ours
Ant	0.608	0.632	0.635	0.612	0.556	0.616	0.600	0.632	0.742	0.800
Camel	0.528	0.545	0.540	0.543	0.524	0.548	0.549	0.616	0.632	0.678
Forrest	0.457	0.627	0.499	0.540	0.544	0.604	0.552	0.579	0.487	0.480
Ivy	0.619	0.620	0.615	0.611	0.563	0.620	0.621	0.635	0.817	0.869
Log4j	0.422	0.493	0.424	0.450	0.475	0.477	0.449	0.512	0.518	0.526
Lucene	0.477	0.570	0.574	0.560	0.533	0.587	0.579	0.490	0.631	0.672
Poi	0.601	0.599	0.563	0.551	0.553	0.589	0.590	0.616	0.671	0.705
Synapse	0.523	0.610	0.606	0.636	0.543	0.602	0.578	0.630	0.702	0.775
Velocity	0.546	0.591	0.577	0.496	0.541	0.595	0.576	0.607	0.655	0.736
Xalan	0.609	0.669	0.640	0.676	0.654	0.707	0.660	0.624	0.687	0.743
Xerces	0.534	0.594	0.573	0.561	0.531	0.576	0.541	0.631	0.634	0.689
Average	0.539	0.595	0.568	0.567	0.547	0.593	0.572	0.597	0.652	0.698
W (T) (L)	11/0/0	10/0/1	10/0/1	10/0/1	10/0/1	10/0/1	10/0/1	10/0/1	10/0/1	—
Improvement	29.5%	17.3%	22.9%	23.1%	27.6%	17.8%	21.9%	16.8%	7.1%	—
p-Value	0.003	0.026	0.004	0.004	0.006	0.016	0.004	0.013	0.004	—