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

Table 5

LED dataset, 100 000 examples. Seven change points: every 12 500 examples. Columns A, = 0 and columns B, = 1000.

Classifiers	LED				SEA
	Final accuracy (%)		Time (s)		Final accuracy (%)		Time (s)
	A	B	A	B	A	B	A	B
FAE	72,27	71,95	74,69	81,32	87,46	86,95	11,37	10,67
AUE	71,52	71,59	133,54	120,84	85,99	86,4	15,09	14,96
AWE	70,38	70,06	236,58	210,79	86,58	86,5	30,62	30,17
OzaBagAdwin	71,68	72,08	86,67	91,37	86,13	86,31	12,99	13,23
Hoeffding Tree	62,65	62,71	8,5	7,92	84,31	84,35	1,09	1,19