Objective To examine the feasibility of using multiple seasonal autoregressive integrated moving average (SARIMA) model for predicting pulmonary tuberculosis (TB) incidence, so as to provide scientific evidence for the prevention and treatment of TB. Methods EViews 7.0.0.1 software was used to create a SARIMA fit model for seasonal incidence of TB on a monthly basis from January 2004 to December 2012, and the predicting performance of the model was tested with TB data from January to December in 2013. Results The established SARIMA (2,0,2)×(0,1,1)₁₂ model could better fit with the previous TB incidence; and it basically well predicted the TB incidence of the 12 months of 2013, with the mean absolute error being 0.416 992 and the mean absolute error rate being 5.350 8%. Conclusion The established multiplicative SARIMA model can better simulate and predict the trend of TB incidence with time, and it may have a future in predicting the incidence of TB.