Table 2. Pearson correlations with the ground-truth data per country.

	NL	BE	IT	FR	UK	ES	PT	DK	IE
(i) IN_NMF vs IN_ECDC for the seasons 2011-2017
	0.91	0.92	0.86	0.83	0.92	0.86	0.84	0.90	0.82
(ii) IN_NMF vs GP for the seasons 2011-2017
	0.88	0.80	0.69	0.79	0.74	0.65	0.66	0.71	0.38
(iii) IN_ECDC vs GP for the seasons 2011-2017
	0.79	0.72	0.80	0.86	0.75	0.67	0.63	0.68	0.23
(iv) IN_NMF forecast vs GP for the season 2016-2017
	0.85	0.82	0.69	0.80	0.60	0.84	0.80	0.76	0.60
(v) IN_NMF forecast vs IN_ECDC for the season 2016-2017
	0.85	0.82	0.86	0.93	0.67	0.59	0.88	0.80	0.71

(i) Pearson correlation between the time series of IN_NMF with the respective time series produced when applying the ILI definition on the Influenzanet data (IN_ECDC). (ii) Pearson correlation between IN_NMF and the respective ILI incidence reported by the national surveillance systems per country (GP). (iii) Pearson correlation between ILI incidence obtained by applying the ECDC case definition to raw Influenzanet data (IN_ECDC) and ILI incidence reported by the national surveillance systems per country (GP). (iv) Pearson correlation between the forecasted 2016-2017 IN_NMF and ILI incidence reported by the national surveillance systems per country (GP) for the season 2016-2017. (v) Pearson correlation between ILI incidence obtained by applying the ECDC case definition to raw Influenzanet data (IN_ECDC) and the respective forecasted IN_NMF for the 2016-2017. Note that the reported correlations are not averages per ILI seasons per country but the correlation of the time series of the entire period (2011-2017 for (i),(ii) and (iii) and 2016-2017 for (iv) and (v)) between the IN_NMF and the respective GP time series for each country.