Fast Adapting Ensemble: A New Algorithm for Mining Data Streams with Concept Drift

Table 2

LED concept, 100 000 examples. One change point: . Columns A, ; columns B, ; columns C, ; and columns D, .

Classifiers	Lowest accuracy around change point (%)				Final accuracy (%)				Time (s)
	A	B	C	D	A	B	C	D	A	B	C	D
FAE	45,1	50,9	55,1	54,1	73,15	73,2	73,16	73,15	63,01	64,55	66,75	68,09
AUE	19,9	15,6	26,6	31	72,84	72,99	72,95	72,89	128,12	118,81	109,25	125,49
AWE	45,2	44,7	27,5	29,7	73,13	73,12	72,97	72,87	234,52	238,35	230,43	204,13
OzaBagAdwin	33,6	61,6	61,1	57,8	72,68	73,53	73,48	73,44	101,57	102,13	101,43	102,68
Hoeffding Tree	16,9	18	21,7	26,7	69,24	69,24	69,26	69,24	9,05	8,95	8,36	9,05