Abstract: Objective Through exploring the use of the moving average method and gray residual GM model to establish prediction early warning model and predict the epidemic tendency of bacillary dysentery in Guangdong Province, to provide scientific database for future prevention and control. Methods Bacillary dysentery incidence data in 2004-2011 in Guangdong Province were collected. We established Gray residual GM Model based on the moving average method (Type III) and adopted the average relative error, posterior difference, and small error probability as the test indicators to compare and analyze the traditional gray GM model (Type I) and Gray GM model based on the moving average method (Type II ). Results The results showed that: C=0.22, P =1.00, the accuracy of the model(Type III) is 1 (good), the average relative error is 5.21%. The model level and average relative error of Model (Type III) is relatively small compared with the other two models (types I and II). The estimates of incidence of bacillary dysentery in Guangdong Province in 2012, 2013, and 2014 using model (Type III) are 5.069 3/100 000, 4.514 0/100 000 and 4.027 0/100 000, respectively. Conclusion Gray residual GM model based on the moving average method (Type III) has the highest fitting accuracy compared with the other two models. It provides the evidence of feasibility and effectiveness to predict the epidemic tendency of bacillary dysentery in Guangdong Province.