Dairy Science and Management

Table 5 Predictive capacity of algorithms fit using data undersampled to balance across classes in sampled data

From: Open-source carbon dioxide and volatile organic compound sensing and associations with defecation and urination events in horses

	Random Forest			Support Vector Machine			Extreme Gradient
	Other	Def^a	Urine^a	Other	Def^a	Urine^a	Other	Def^a	Urine^a
Predicted
Other	1768	2	2	2449	14	2	1281	4	0
Def^a	2465	32	3	1748	16	3	3241	31	6
Urine^a	1739	6	2	1775	10	2	1450	5	1
Class Statistics
Sensitivity	0.29	0.80	0.29	0.41	0.40	0.29	0.21	0.78	0.14
Specificity	0.91	0.59	0.71	0.66	0.71	0.70	0.91	0.46	0.76
PPV^a	1.00	0.01	0.001	0.994	0.009	0.001	0.997	0.009	0.001
NPV^a	0.01	0.998	0.999	0.009	0.994	0.999	0.009	0.997	0.999
Accuracy	0.61	0.69	0.50	0.53	0.55	0.49	0.56	0.62	0.45
Overall Statistics
Accuracy	0.299			0.410			0218
NIR^a	0.992			0.992			0.992

^aDefecation (Def) and urination (Urine) events, positive predictive value (PPV), negative predictive value (NPV), and null information rate (NIR)

Back to article page

ISSN: 3005-0707

Contact us

General enquiries: info@biomedcentral.com