Objective To compare the accuracy of the seasonal time series decomposition method and autoregressive integrated moving average (ARIMA) in the prediction of incidence of tuberculosis(TB) in order to facilitate early-warning.Methods The seasonal decomposition model and ARIMA model were constructed by SPSS20.0 software based on time series of monthly TB incidence between January 2005 and December 2014 in Urumqi,China.The obtained models were used to forecast the monthly incidence in 2015 and compared with the actual incidence respectively.Results Between 2005 and 2014,the incidence of TB was higher during March,April and May in Urumqi.A linear fitting model and a cubic curve fitting model were constructed by the time series seasonal decomposition method.The mean absolute percentage error (MAPE) of each predicted monthly incidence in 2015 was 18.75% and 92.25%,respectively.The predicted values of the linear model were lower than actual values and the predicted values of the cubic curve model were higher than actual values.An ARIMA (2,1,1) (1,1,0)12 fitting model was established by ARIMA method.The MAPE of each predicted monthly incidence in 2015 was 9.46% and there were no significant differences between the predicted and actual values.Conclusion The ARIMA method is better than the seasonal decomposition method for predicting the monthly incidence of TB in Urumqi.