Using electronic health records and Internet search information for accurate influenza forecasting

BMC Infectious Diseases

Table 2 ARGO performance compared to alternative methods for the validation period of February 28, 2015 to July 2, 2016

	real-time	forecast 1 week	forecast 2 week	forecast 3 week
RMSE
ARGO	0.341	0.540	0.604	0.704
healthmap	0.530	0.590	0.932	0.949
ar4	0.902	0.909	0.838	0.780
naive	1 (0.206)	1 (0.330)	1 (0.439)	1 (0.552)
MAE
ARGO	0.386	0.502	0.529	0.563
healthmap	0.527	0.564	0.697	0.700
ar4	0.994	0.952	0.852	0.766
naive	1 (0.146)	1 (0.248)	1 (0.341)	1 (0.435)
RMSPE
ARGO	0.425	0.472	0.524	0.593
healthmap	0.622	0.613	0.868	0.871
ar4	0.959	1.006	0.958	0.920
naive	1 (0.108)	1 (0.173)	1 (0.232)	1 (0.293)
MAPE
ARGO	0.448	0.466	0.489	0.494
healthmap	0.592	0.593	0.666	0.654
ar4	1.034	1.018	0.935	0.860
naive	1 (0.083)	1 (0.139)	1 (0.194)	1 (0.250)
Correlation
ARGO	0.995	0.963	0.916	0.823
healthmap	0.987	0.956	0.843	0.774
ar4	0.961	0.896	0.842	0.776
naive	0.963	0.900	0.829	0.745
Error reduction of ARGO over the best alternative (in %)
RMSE	35.63	8.38	27.94	9.77
MAE	26.75	11.07	24.16	19.49
RMSPE	31.63	22.94	39.59	31.93
MAPE	24.29	21.42	26.58	24.42

The evaluation metrics are defined in Table 1. The benchmark methods are the same as Table 1 except that the ensemble method of Santillana et al. [11] is replaced by a refined version broadcasted by the Healthmap Flu Trends system. Boldface highlights the best method for each metric in each forecasting time horizon. RMSE, MAE, RMSPE, MAPE are relative to the error of the naive method, i.e., the numbers are the ratio of the error of a given method over that of the naive method; the absolute error of the naive method is given in the round bracket. Table S4 in the Additional file 1 gives absolute error of all methods. For each forecasting time horizon and each evaluation metrics, the error reduction of ARGO over the best alternative method is given in the second half of the table.

ISSN: 1471-2334