Performance metrics for best neural network models compared with Dante.

Forecasting performance metrics for the best-performing neural network (SRNN for γ = 7, IRNN for γ ≥ 14) compared with Dante (Dte) and Elastic Net (Eln). The NNs are trained using search query frequencies generated only up to the last available ILI rate (the 2-week advantage of using Web search data is removed). We use leave-one flu season-out to train models, similarly to Dante. The best results for this comparison are shown in bold. NNb denotes results where the temporal advantage of Web search activity information is maintained (compared to NN). NNa holds results for the same experiment as NNb with the addition of disabling leave-one flu season-out training, i.e. training does not include data after the test year. Eln uses the same data sets (inputs, targets) as the NNs. Therefore, it is trained using look ahead and without leave-one flu season-out. Eln does not estimate uncertainty and hence, the Skill metric is not available (empty cell). This Table supplements Table 2 in the main manuscript. (XLSX)