Balanced Adversarial Tight Matching for Cross-Project Defect Prediction

Table 4

The results of the measure comparison with nine methods.

Target project	LR	NNFilter	TCA	TCA+	DBN	DPCNN	TCNN	MANN	ADA	Ours
Ant	0.445	0.443	0.444	0.418	0.367	0.425	0.410	0.515	0.506	0.547
Camel	0.328	0.325	0.326	0.328	0.310	0.333	0.335	0.405	0.401	0.398
Forrest	0.192	0.167	0.114	0.120	0.131	0.152	0.133	0.229	0.256	0.274
Ivy	0.289	0.274	0.258	0.250	0.229	0.268	0.268	0.254	0.332	0.455
Log4j	0.435	0.637	0.661	0.650	0.654	0.659	0.686	0.467	0.594	0.705
Lucene	0.603	0.597	0.619	0.551	0.584	0.641	0.632	0.605	0.543	0.615
Poi	0.508	0.634	0.605	0.589	0.612	0.622	0.669	0.634	0.512	0.506
Synapse	0.525	0.520	0.522	0.549	0.453	0.520	0.495	0.584	0.569	0.560
Velocity	0.506	0.501	0.496	0.310	0.453	0.517	0.500	0.515	0.461	0.518
Xalan	0.534	0.605	0.647	0.655	0.651	0.625	0.680	0.572	0.588	0.622
Xerces	0.532	0.611	0.607	0.577	0.590	0.605	0.620	0.584	0.587	0.669
Average	0.445	0.483	0.482	0.455	0.458	0.487	0.494	0.488	0.486	0.533
W (T) (L)	10/0/1	10/0/1	8/0/3	9/0/2	9/0/2	8/0/3	8/0/3	8/0/3	8/0/3	—
Improvement	19.8%	10.3%	10.7%	17.3%	16.5%	9.4%	8.0%	9.3%	9.6%	—
p-Value	0.004	0.041	0.042	0.026	0.026	0.045	0.048	0.047	0.016	—