Feature Selections Using Minimal Redundancy Maximal Relevance Algorithm for Human Activity Recognition in Smart Home Environments

Table 11

Optimal recognition accuracy rates under three evaluation criteria (MR, D-R, and D/R) and the required minimal size of features subset (in brackets).

Activity
Classifier	Criterion	0	1	2	3	4	5	6	7	8	9	Total
NB	MR	0.667 (12)	1.000 (8)	0.966 (13)	0.739 (10)	1.000 (1)	0.300 (3)	0.986 (10)	1.000 (1)	0.881 (1)	0.909 (9)	0.903 (12)
	D-R	0.667 (6)	0.979 (2)	0.966 (13)	0.848 (2)	1.000 (1)	0.600 (4)	0.986 (9)	1.000 (1)	0.881 (1)	0.886 (6)	0.903 (9)
	D/R	0.733 (9)	0.979 (2)	0.966 (13)	0.848 (2)	1.000 (1)	0.600 (4)	0.986 (11)	1.000 (1)	0.881 (1)	0.864 (3)	0.903 (11)
HMM	MR	0.900 (1)	0.938 (6)	0.961 (13)	0.804 (4)	1.000 (10)	0.400 (10)	1.000 (1)	0.973 (5)	0.821 (13)	0.932 (5)	0.880 (13)
	D-R	0.900 (1)	0.792(10)	0.961 (13)	0.978 (5)	1.000 (7)	0.700 (4)	1.000 (1)	0.973 (6)	0.821 (13)	0.955 (9)	0.880 (13)
	D/R	0.900 (1)	0.750 (12)	0.961 (13)	0.978 (5)	0.976 (12)	0.700 (4)	1.000 (1)	0.973 (12)	0.821 (13)	0.955 (8)	0.880 (13)