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

Table 4

LED and SEA concepts, 100 000 examples. Three change points: every 25 000 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	72,65	72,36	83,9	82,98	87,52	87,48	14,6	11,72
AUE	71,82	71,79	142,23	143,58	87,51	87,23	15,12	15,23
AWE	72,43	71,52	233,14	239,77	87,3	87,17	31	30,84
OzaBagAdwin	71,39	72,6	99,23	84,26	86,97	87,21	13,29	13,42
Hoeffding Tree	61,22	61,79	8,07	5,32	85,61	85,6	1,05	1,15