Super machine learning: improving accuracy and reducing variance of behaviour classification from accelerometry

Table 4 Confusion matrices from three test models using four behaviours and 13 epochs

	Foraging	Grooming	Resting	Travelling	Precision	Sensitivity
Super learner
Foraging	1248	17	53	182	0.83	0.82
Grooming	18	1292	27	64	0.92	0.91
Resting	80	37	1321	61	0.88	0.89
Travelling	185	79	77	1158	0.77	0.79
Gradient boosting machine
Foraging	1243	23	54	180	0.83	0.81
Grooming	20	1300	30	52	0.93	0.89
Resting	80	39	1305	76	0.87	0.90
Travelling	191	92	68	1149	0.77	0.79
Random forest
Foraging	1220	25	57	198	0.81	0.80
Grooming	17	1291	42	52	0.92	0.90
Resting	86	35	1312	67	0.87	0.89
Travelling	195	88	59	1158	0.77	0.79

ISSN: 2050-3385