Analysis and Prediction of Pulmonary Tuberculosis Using an ARIMA Model in Shaanxi Province, China

Abstract An analysis and prediction for the incidence of tuberculosis (TB) is particularly important since TB still has a high fatality rate in the world. However, this prediction is often influenced by inaccurate forecasting ways. We used data from 364,762 reported TB cases between January 2005 and December 2015 in Shaanxi Province, China. The known number of cases in 2016 was used to assess the accuracy of the model’s predictions. Through all aspects of analysis and comparison, the ARIMA (0, 1, 2) (0, 1, 1)12 were the most model. In the fitting dataset, for the ARIMA (0, 1, 2) (0, 1, 1)12 model, RMSE, MAPE, MAE and MER were 0.7667, 6.7810, 6.04944 and 0.06836, respectively; And in the forecasting dataset were 0.32808, 6.01834, 0.2899 and 0.0615, respectively. The model can predict the seasonal changes and trends of tuberculosis in the Shaanxi province’s population.