Table 3.

10-fold cross-validated SAYL performance under various parameters. Minpos is the minimum number of positive examples that a rule is required to cover. θ is the up-lift AUC improvement threshold for adding a rule to the theory. We also include results of SAYL using cross-validated parameters and Differential Prediction Search (DPS).

minpos	θ (%)	search mode	Uplift AUC	Lift(older) AUC	Lift(younger) AUC	Rules Avg #	DPS p-value
13	1	g-SAYL	63.29	96.79	33.50	16.4	0.002	*
13	1	e-SAYL	43.51	83.82	40.31	2.0	0.049	*
13	5	g-SAYL	58.06	96.14	38.07	5.9	0.002	*
13	5	e-SAYL	53.37	85.66	32.29	1.8	0.027	*
13	10	g-SAYL	61.68	96.26	34.58	3.6	0.002	*
13	10	e-SAYL	65.36	90.50	25.14	1.1	0.002	*
7	1	g-SAYL	65.48	98.82	33.34	18.3	0.002	*
7	1	e-SAYL	25.50	74.39	48.90	3.0	0.695
7	5	g-SAYL	58.91	96.67	37.76	5.8	0.002	*
7	5	e-SAYL	32.71	79.52	46.81	2.5	0.557
7	10	g-SAYL	61.98	96.87	34.89	3.6	0.002	*
7	10	e-SAYL	52.35	83.64	31.29	1.6	0.002	*
-	-	SAYL	58.10	97.24	39.15	9.3	0.002	*
13	-	DPS	27.83	101.01	73.17	37.1	-

We compute the p-value comparing each method to DPS, * indicating significance.