Objective: To investigate the application of bootstrap method in the early warning for influenza surveillance with 95%CIs of weekly influenze-like illnesses(ILI)% estimated by bootstrap method specified as the early warning thresholds. Methods: The influenza surveillance data from 2005 to 2014(not including 2006 and 2009) were taken as baseline. Thirty Percent of the highest positive rate of influenza viruses in two years was taken as the gold standard of influenza pandemic. Consistency of the early warning results was compared among bootstrap method, moving average method and the gold standard. We also discussed the application of bootstrap in practice in 2016 based on the surveillance data from 2012 to 2015. Results: Influenza viruses circulated throughout the year and peaked during cold winter and spring season from November to following February and hot summer from June to August. The result of bootstrap method has high consistency with moving average method and gold standard. The bootstrap method was more sensitive in the early warning of influenza pandemic in 2009 compared with moving average method and more consistent with gold standard(kappa=0.728). Conclusion: Bootstrap was a compute-intensive method which was not rely on the distribution of ILI% data, and it can be applied as an alternative and complementary method in the early warning for influenza surveillance.