Comparison of new computational methods for spatial modelling of malaria

Malaria Journal

Table 4 Cross-validation results grouped by density of sampled locations

Fold	Density	Model	RMSE	Corr	% points with absolute error less than
Fold	Density	Model	RMSE	Corr	0.05	0.1	0.2
10-fold	Low	INLA	0.005	− 0.234	100	100	100
		GPBoost	0.028	− 0.189	96	100	100
		SpRF	0.006	− 0.059	100	100	100
		FRK	0.048	− 0.021	62	98	100
	Medium	INLA	0.141	− 0.003	75.723	82.659	93.642
		GPBoost	0.138	− 0.021	60.694	85.549	94.22
		SpRF	0.143	− 0.025	72.832	80.347	93.642
		FRK	0.145	− 0.035	58.382	76.301	94.22
	High	INLA	0.24	0.305	52.229	61.783	73.248
		GPBoost	0.134	0.788	29.936	52.866	89.809
		SpRF	0.14	0.737	56.051	73.885	85.35
		FRK	0.119	0.828	54.14	70.064	87.261
50-fold	Low	INLA	0.005	− 0.128	100	100	100
		GPBoost	0.017	− 0.202	98	100	100
		SpRF	0.006	− 0.077	100	100	100
		FRK	0.021	− 0.17	96	100	100
	Medium	INLA	0.123	0.501	73.41	83.815	90.173
		GPBoost	0.117	0.511	73.41	84.393	91.908
		SpRF	0.121	0.435	71.676	85.549	96.532
		FRK	0.133	0.31	75.723	83.237	91.329
	High	INLA	0.143	0.776	55.414	71.338	81.529
		GPBoost	0.119	0.809	47.134	70.701	84.713
		SpRF	0.14	0.742	52.866	64.968	80.892
		FRK	0.129	0.784	45.86	68.153	85.987

Boldface denotes the best RMSE score for a given number of cross validation folds and point density

ISSN: 1475-2875