Objective To explore the feasibility of Sparse coefficient autoregressive integrated moving average( ARIMA) model for predicting tuberculosis incidence and provide a scientific evidence for the targeted prevention and control policy of TB. Methods Sparse coefficient ARIMA model was built using tuberculosis surveillance data from January 2005 to December 2013 in Jiangxi province, and the predictive effect was evaluated by using the data from 2014 to 2015. Results The annual seasonal effect in the incidence of tuberculosis was observed from January 2005 to December 2013 in the province, and the long-term descending trend in the incidence of tuberculosis was also observed. Sparse coefficient ARIMA (0,(1,12),(2,3)) model could better fit the incidence of tuberculosis over the period, and forecast situation and actual situation were basically similar. Conclusion Sparse coefficient ARIMA model could effectively clarify the regularity of the incidence of tuberculosis in Jiangxi province and predict the trend of the incidence.