Comparison of the effectiveness of five time series models for prediction ofpulmonary tuberculosis incidence
WANG Yingdan1, GAO Chunjie1, WANG Lei2
1. School of Public Health, Xinjiang Medical University, Urumqi, Xinjiang 830011, China; 2. Xinjiang Medical University, Urumqi, Xinjiang 830011, China
Abstract：Objective To compare the effectiveness of seasonal autoregressive integrated moving average (SARIMA) model, additive Holt-Winters model, Holt-Winters' multiplicative model, GM (1, 1) model and linear combination prediction model for prediction of pulmonary tuberculosis incidence. Methods Data pertaining to monthly incidence of pulmonary tuberculosis in Xinjiang Uyghur Autonomous Region from 2004 to 2008 were captured from Public Health Sciences Data Center. The SARIMA model, additive Holt-Winters model, Holt-Winters' multiplicative model, GM (1, 1) model and linear combination prediction model were created based on the incidence of pulmonary tuberculosis from January 2004 to June 2018, to predict the incidence of pulmonary tuberculosis from July to December 2018. The predictive value of each model was evaluated using absolute percentage error (APE), mean APE (MAPE) and root mean square error (RMSE), and the best model was selected based on minimum APE, MAPE and RMSE. Results The SARIMA model showed the minimum APE (10.94%), 11.01% and 7.96% MAPE and 564 and 419 RMSE at the model-fitting and prediction phases; followed by the linear combination prediction model, with 13.71% APE, 12.01% and 7.94% MAPE and 600 and 447 RMSE at the model-fitting and prediction phases, while the additive Holt-Winters model, Holt-Winters' multiplicative model and GM (1, 1) model showed a low predictive value. Conclusion The SARIMA and linear combination prediction models are superior to additive Holt-Winters model, Holt-Winters' multiplicative model and GM (1, 1) model for prediction of pulmonary tuberculosis incidence.
王迎丹, 高春洁, 王蕾. 5种时间序列模型预测肺结核发病比较[J]. 预防医学, 2022, 34(12): 1194-1200.
WANG Yingdan, GAO Chunjie, WANG Lei. Comparison of the effectiveness of five time series models for prediction ofpulmonary tuberculosis incidence. Preventive Medicine, 2022, 34(12): 1194-1200.
[1] BROWN J S,BATESON T F,MCDONNELL W F.Effects of exposure to 0.06 ppm ozone on FEV1 in humans:a secondary analysis of existing data[J].Environ Health Perspect,2008,116(8):1023-1026. [2] 赵棋锋,马珊珊,王吉玲,等.指数平滑法与ARIMA模型对流感样病例流行趋势的预测效果比较[J].预防医学,2020,32(4):381-383,387. ZHAO Q F,MA S S,WANG J L,et al.Comparison of exponential smoothing method and ARIMA model on epidemic trends in influenza-like cases[J].Prev Med,2020,32(4):381-383,387. [3] 孙振球. 医学统计学[M].北京:人民卫生出版社,2005:461. SUN Z Q.Medical Statistics[M].Beijing:People's Medical Publishing House,2005:461. [4] 徐国祥. 统计预测和决策[M].上海:上海财经大学出版社,2012:164. XU G X.Statistical forecasting and decision making[M].Shanghai:Shanghai University of Finance and Economics Press,2012:164. [5] 刘超,孟园园,张庆雯.手足口病发病预测4种时间序列预测模型比较[J].中国公共卫生,2022,38(2):218-223. LIU C,MENG Y Y,ZHANG Q W.Comparison of four time series prediction models for the incidence of hand-foot-and-mouth disease[J].Chin J Public Health,2022,38(2):218-223. [6] 竺盛波,沈妙儿,周聪盛,等.奉化区缺血性脑卒中发病季节的SARIMA模型预测[J].预防医学,2018,30(2):176-178,181. ZHU S B,SHEN M E,ZHOU C S,et al.SARIMA model prediction of ischemic stroke season in Fenghua area[J].Prev Med,2018,30(2):176-178,181. [7] 傅颖,张晓龙,蒋骏,等.SARIMA模型在苏州市肺结核发病预测中的应用[J].中国热带医学,2020,20(4):339-342. FU Y,ZHANG X L,JIANG J,et al.Application of SARIMA model in the prediction of tuberculosis incidence in Suzhou[J].Chin J Trop Med,2020,20(4):339-342. [8] 张鲁玉,孙亮,马兰,等.SARIMA模型和Holt-Winters模型在我国丙肝月报告发病人数预测中的应用比较[J].现代预防医学,2020,47(21):3855-3858,3951. ZHANG L Y,SUN L,MA L,et al.Comparison of SARIMA model and Holt-winters model in the prediction of the number of reported cases of hepatitis C in China[J].Mod Prev Med,2020,47(21):3855-3858,3951. [9] 汪鹏,彭颖,杨小兵.ARIMA模型与Holt-Winters指数平滑模型在武汉市流感样病例预测中的应用[J].现代预防医学,2018,45(3):385-389. WANG P,PENG Y,YANG X B.Application of ARIMA model and Holt-Winters exponential smoothing model in the prediction of influenza-like cases in Wuhan[J].Mod Prev Med,2018,45(3):385-389. [10] 赵棋锋,马珊珊,王吉玲,等.指数平滑法与ARIMA模型对流感样病例流行趋势的预测效果比较[J].预防医学,2020,32(4):381-383,387. ZHAO Q F,MA S S,WANG J L,et al.Comparison of the effect of exponential smoothing method and ARIMA model on the prediction trend of influenza-like cases[J].Prev Med,2020,32(4):381-383,387. [11] 刘天,王芸,姚梦雷,等.GM (1, 1) 灰色模型、马尔可夫链模型及其组合模型和SARIMA模型在甲肝发病数预测中的应用效果比较[J].华南预防医学,2019,45(2):128-132. LIU T,WANG Y,YAO M L,et al.Comparison of the application effect of GM (1, 1) grey model,Markov chain model and its combination model and SARIMA model in predicting the incidence of hepatitis A[J].South China J Prev Med,2019,45(2):128-132. [12] 官陈平,傅发源,王依妹,等.灰色预测模型GM (1, 1)在肺结核发病趋势预测中的应用[J].医学动物防制,2021,37(10):931-935. GUAN C P,FU F Y,WANG Y M,et al.Application of grey predictive model GM (1, 1) in prediction of pulmonary tuberculosis incidence trend[J].Med Anim Control,2021,37(10):931-935. [13] 张伟文,贺湘焱,古丽娜扎尔·艾克拜尔,等.季节时间序列分析在新疆地区涂阳结核疫情预测预警中的应用[J].实用预防医学,2019,26(1):26-29. [14] 王华,田昌伟,王文明,等.应用SARIMA-GRNN组合模型分析肺结核流行的季节性特征[J].预防医学,2019,31(1):55-58. WANG H,TIAN C W,WANG W M,et al.Application of SARIMA-GRNN combination model to analyze the seasonal characteristics of tuberculosis epidemic[J].Prev Med,2019,31(1):55-58.
