Fig 1.
The flow chart for this study.
Fig 2.
An example of cross-correlation analysis.
Table 1.
Maximum correlation coefficient and time lag with time series of influenza surveillance data in the United States and input variables.
Fig 3.
The explanation for forecasting seasonal influenza after 26 weeks.
Fig 4.
The explanation for the training set at the 40th week and output for forecasting seasonal influenza after 26 weeks.
Fig 5.
The surveillance data for influenza A (a) and B (b) viruses in the U.S. and Australia; the values for Australia were shifted forward 22 weeks in 2010–2018. The surveillance data for influenza A (c) viruses in the U.S. and Chile; the values for Australia were shifted forward 28 weeks. The blank part of the graph, the gap between INF A and the sum of the H1N1 and H3 viruses, is the number of influenza viruses (not subtyped).
Table 2.
Linear regression analysis for influenza activities of previous season in the U.S. and input variables from the 40th week in 2010 to the 40th week in 2018.
Table 3.
Performance of prediction models for seasonal influenza outbreaks after 26 weeks in the United States from 2015 to 2019.
Fig 6.
The prediction of ANN for ILI (a) and Total influenza (b), influenza A and influenza B viruses (c) after 26 weeks from 2015 to 2019 in the U.S.
Fig 7.
The prediction and 95% confidence interval of Auto Regressive Integrated Moving Average for ILI (a) and Total influenza (b), influenza A (c) and influenza B viruses (d) from 2015 to 2019 in the U.S.