Application of three models to predict the influenza activity of 2011 -2012

Objective To construct a proper model to predict the peak of influenza activity in 2011 -2012 flu season.Methods Influenza virus surveillance information in flu seasons from 2007 to 2010,collected by etiological surveillance of influenza in Beijing,were selected and used as the baseline data.Grey Model ( GM( 1,1 ) ),autoregressive equation(AR) and binomial equation were developed and validated based on the ratio of posterior deviation and coefficient of determination.The selected model and parameters were used to predict the influenza activity in 2011- 2012 flu season.Results GM (1,1) predictionequation was Y,=38 -37.78 e-0.01t where C =0.89 with pcor fitting.AR(1) equation was Yi =0.07 + 0.77 Yi-1 where R2 =0.58.The equation was statistically significant (P ＜0.001 ).AR(2) equation was Yi =0.09 + 0.94 Yi-1 -0.23 Yi-2 where R2 =0.58.The equation was statistically insignificant (P=0.24).The binomial equation was Y =-0.06x2 + 1.95x + 22.46 where R2 =0.92.The equation was regarded as the best model to predict the peak activity in 2011 - 2012 season.The predictive peak of positive rate would be appearing in the 48th week of 2011,and the maximum positive rate was expected as 37.67％.Conclusion The binomial equation was the most appropriate model in predicting the intensity of influenza activity. Key words: influenza; Gray model; autoregressive equation; binomial equation; prediction