Abstract: Objective To analyze the correlation between meteorological factors and the incidence of malaria in Wanning municipality (Wanning) of Hainan province,and to establish back propagation(BP)neural network model and stepwise regression model of malaria incidence and then to evaluate the predictive effect of the two models.Methods Monthly meteorological data and incidences of malaria in Wanning from January 1995 through December 2007 were collected;Spearman's rank correlation was applied to analyze the association between the meteorological factors and the incidences of malaria.BP neural network method and stepwise regression method were adopted to establish fitting models of malaria incidence by introduced meteorological parameters using Matlab 7.0 and SPSS software;monthly malaria incidence in 2008 was estimated with the two models constructed.Results The incidence of malaria in Wanning was positively correlated with average air temperature,maximum temperature,minimum temperature,rainfall,and sunshine time in the previous month (P<0.05 for all),and negatively correlated with average relative humidity and air pressure (both P<0.01).A BP neural network model containing a hidden layer was established with 7 meteorological factors as the input variables and the incidence of malaria as the output variable;the model with 16 hidden units presented the best fitting,with the minimum training error of 0.001 after 300 training epochs;the mean square error and determination coefficient (R²) of the model were 0.0027 and 0.9900,respectively.With the same 7 input variables and the one output variable,a stepwise regression model was also established and average air temperature and relative humidity were introduced into the model,and the R² of the model was 0.40.The monthly malaria incidence in Wanning were predicted using the two models established and the mean absolute error of the prediction was 1.24/10 000 and 0.44/10 000,respectively.Conclusion There is a significant correlation between the incidence of malaria and the meteorological factors in Wanning.The fitting efficiency of BP neural network model established with meteorological factors is higher than that of stepwise regression model,but the prediction efficiency is relatively lower,suggesting the generalization capacity of the BP neural network model needs to be improved.