Table 2.

Cross-validation and independent test scores of random forest regression model for predicting small ruminants (sheep & goat) density with different buffer distances and data split ratios for 2015 (lower RMSE/MAE and higher R² indicate better model fit).

Buffer distance	Cross-Validation score			Independent test score
	RMSE	R2	MAE	RMSE	R2	MAE
same location
70-30	70	0.49	17.29	84.17	0.48	20.93
80-20	65.3	0.59	16.74	87.21	0.29	19.29
90-10	68.96	0.63	17.08	50.29	0.21	14
2 km buffer
70-30	52.22	0.51	13.4	62.96	0.7	12.35
80-20	61.09	0.51	14.46	22.98	0.59	8.65
90-10	57.13	0.54	13.13	20.33	0.68	8.38
5 km buffer
70-30	49.55	0.6	11.26	30.2	0.33	7.09
80-20	45.63	0.62	9.9	35.77	0.25	7.9
90-10	44.07	0.6	9.3	35.67	0.38	7.22
10 km buffer
70-30	21.11	0.62	5.84	11.13	0.79	4.74
80-20	19.45	0.67	5.4	12.68	0.73	4.94
90-10	17.53	0.76	4.87	11.75	0.82	4.74

The Cross-Validation and Independent test scores were computed between the observed (census data) and predicted (RF model) livestock density values using the training and testing datasets, respectively.