TABLE 4.

Spam data. From left to right: the recall on test data under various cost ratios, the precision corresponding to the highest recall, training time, and test time.

	Recall with different cost ratios C +: C −							Precision	Training Time(s)	Test Time(s)
	1:1	1.5:1	2:1	2.5:1	5:1	7.5	10:1
Adacost	0.633	0.669	0.698	0.878	0.698	0.92	0.775	0.1	1.41	0.004
AdaC1	0.633	0.824	0.885	0.896	0.917	0.99	0.99	0.076	3.51	0.004
AdaC2	0.633	0.9	0.92	0.96	0.98	0.99	0.99	0.076	2.73	0.005
AdaC3	0.594	0.881	0.94	0.824	0.824	0.775	0.91	0.33	2.1	0.004
DiffBoost	0.931	0.931	0.931	0.931	0.931	0.931	0.931	0.392	1.35	0.003
AdaClassWeight	0.92	0.92	0.92	0.92	0.92	0.92	0.92	0.403	0.74	0.003