Table 2.

Accuracy, precision and recall values obtained for the 4 behaviour categories, for each algorithm. The weighted average across behaviour categories is also given

Algorithm	Classification performance	Running	Walking	Digging	Motionless	Weighted average
Three nearest neighbours	Accuracy	98.18	96.82	97.12	97.58	97.53
	Precision	95.27	91.34	*80.00	98.21	94.90
	Recall	96.58	92.06	81.63	97.05	94.85
Linear support-vector machine	Accuracy	96.36	95.30	94.39	96.67	96.17
	Precision	93.57	87.40	*60.00	97.60	91.97
	Recall	89.73	88.10	73.47	95.87	91.36
Radial basis function kernel SVM	Accuracy	97.12	96.82	96.67	96.97	96.95
	Precision	93.79	92.68	*75.47	97.05	93.89
	Recall	93.15	90.48	81.63	97.05	93.79
Decision tree	Accuracy	97.27	96.21	95.15	97.42	96.99
	Precision	93.84	90.40	*64.41	98.79	93.54
	Recall	93.84	89.68	77.55	96.17	93.03
Random forest	Accuracy	97.58	96.67	97.73	98.03	97.65
	Precision	92.76	90.63	*92.50	97.94	95.00
	Recall	96.58	92.06	75.51	98.23	95.00
Gaussian Naïve Bayes	Accuracy	97.73	96.82	95.61	97.73	97.40
	Precision	93.38	98.17	*65.15	98.50	94.83
	Recall	96.58	84.92	87.76	97.05	93.94
Linear discriminant analysis	Accuracy	98.33	95.91	95.45	97.88	97.42
	Precision	97.20	88.37	*67.27	98.80	94.11
	Recall	95.21	90.48	75.51	97.05	93.79
Artificial neural network	Accuracy	97.42	96.52	96.82	97.42	97.21
	Precision	93.88	89.92	*81.82	97.35	94.01
	Recall	94.52	92.06	73.47	97.64	94.09

Asterisks allow easy comparison of precision across algorithms for digging. The random forest model was retained and is in bold