%0 Journal Article %T 基于SARIMA-LSTM模型的中国肺结核传染病预测研究
Prediction of Tuberculosis Infection in China Based on SARMIA-LSTM Model %A 王晓琴 %A 杨震 %A 包城 %A 郭松柏 %A 许传青 %J Statistics and Applications %P 8-21 %@ 2325-226X %D 2025 %I Hans Publishing %R 10.12677/sa.2025.142030 %X 背景：中国是结核病高负担国家之一，尽管肺结核新发病例数逐年下降，但每年新增感染者的数量一直处于较高水平且肺结核感染者的诊断率较低。目的：选择更精准的预测肺结核的发病情况模型，为肺结核的防控和预警提供科学依据。方法：建立SARIMA和LSTM模型，运用加权组合的方法构建SARIMA-LSTM组合模型，使用平均绝对误差(MAE)、均方根误差(RMSE)和平均绝对误差百分比(MAPE)三个评价指标比较模型的预测性能，确定最佳预测模型，并使用该模型对肺结核发病趋势进行预测。结果：SARIMA模型、LSTM模型和SARIMA-LSTM组合模型的平均绝对百分比误差(MAPE)分别为17.95、14.62、8.49，组合模型的MAPE比SARIMA模型降低了52.70%，比LSTM模型降低了41.89%。结论：SARIMA-LSTM组合模型的拟合效果更好，预测误差在三个模型中最低。该组合模型能发挥单一模型的优势，相比两种单一模型提升了预测的准确性。
Background: China is one of the countries with a high burden of Tuberculosis (TB). Although the number of new cases of TB has been decreasing year by year, the number of newly infected people each year has been at a high level and the diagnosis rate of TB-infected people is low. Objective: To select a more accurate model to predict the incidence of tuberculosis and provide a scientific basis for the prevention and control of tuberculosis and early warning. Methods: The SARIMA and LSTM models were established, and the SARMIA-LSTM combined model was constructed by the weighted combination method. The prediction performance of the model was compared by the three evaluation indexes of Mean Absolute Error (MAE), Root Mean Square Error (RMSE) and Mean Absolute Percentage Error (MAPE), and the optimal prediction model was determined. The model was used to predict the trend of tuberculosis incidence. Results: The Mean Absolute Percentage Error (MAPE) of SARIMA model, LSTM model and SARMI-LSTM combined model were 17.95, 14.62 and 8.49, respectively. The MAPE of the combined model was reduced by 52.70% compared with SARIMA model and 41.89% compared with LSTM model. Conclusion: SARMIA-LSTM combined model has a better fitting effect, and the prediction error is the lowest among the three models. The combined model can give full play to the advantages of a single model and improve the accuracy of prediction compared with the two single models. %K 肺结核， %K SARIMA， %K LSTM
Tuberculosis %K SARIMA %K LSTM %U http://www.hanspub.org/journal/PaperInformation.aspx?PaperID=107477