Adaptive Online Sequential ELM for Concept Drift Tackling

Table 3

Average testing accuracy in % for each concept between OS-ELM, CEOS-ELM, AOS-ELM1, and AOS-ELM2.

(a) Testing accuracy in % for SEA with

Method	Parameter
OS-ELM	NORM	89.48 ± 0.33	86.47 ± 0.37	83.41 ± 0.93	85.32 ± 0.45
	ROS	89.51 ± 0.37	86.43 ± 0.38	83.49 ± 0.95	85.31 ± 0.49
CEOS-ELM	NORM	82.40 ± 0.27	89.33 ± 0.43	75.80 ± 0.87	90.30 ± 0.35
	ROS	84.54 ± 0.59	89.96 ± 0.59	77.97 ± 1.05	90.26 ± 0.69
AOS-ELM1	NORM	89.84 ± 0.28	90.03 ± 0.25	89.67 ± 0.61	90.33 ± 0.39
	ROS	89.76 ± 0.34	90.02 ± 0.25	89.76 ± 0.55	90.34 ± 0.38
AOS-ELM2	NORM	50.58 ± 1.18	50.71 ± 1.19	48.50 ± 10.10	90.34 ± 0.30
	ROS	65.78 ± 1.21	65.67 ± 1.19	64.09 ± 1.89	90.14 ± 1.34

(b) Testing accuracy in % for STAGGER with

Method	Parameter
OS-ELM	NORM	51.89 ± 3.48	81.61 ± 4.74	67.18 ± 5.80
	ROS	49.77 ± 1.96	84.16 ± 1.61	66.93 ± 2.00
CEOS-ELM	NORM	21.98 ± 1.57	53.66 ± 4.34	97.84 ± 4.32
	ROS	23.23 ± 1.93	52.11 ± 2.35	99.27 ± 1.45
AOS-ELM1	NORM	97.64 ± 1.95	100.00 ± 0.00	100.00 ± 0.00
	ROS	100.00 ± 0.00	100.00 ± 0.00	100.00 ± 0.00
AOS-ELM2	NORM	59.66 ± 5.65	70.91 ± 10.93	100.00 ± 0.00
	ROS	56.20 ± 9.56	69.41 ± 14.05	100.00 ± 0.00