Application of SARIMA-GRNN combination model and SARIMA model to predicting the incidence rate of mumps
LIU Tian1,2, YAO Meng-lei1, HUANG Ji-gui1, WU Yang3, CHEN Qi3, TONG Ye-qing3, CHEN Hong-ying3, MEI Fang-sheng1
1. Jingzhou Municipal Center for Disease Control and Prevention, Jingzhou, Hubei 434000, China; 2. Chinese Field Epidemiology Training Program, Beijing 100050, China; 3. Hubei Provincial Center for Disease Control and Prevention, Wuhan, Hubei 430079, China
Abstract:Objective To explore the application of seasonal autoregressive integrated moving average-generalized regression neural network (SARIMA-GRNN) combination model and seasonal autoregressive integrated moving average (SARIMA) model to forecasting the incidence rate of mumps, and to compare the predicated effect between them. Methods Data regarding the monthly incidence rate of mumps in Shanghai from 2004 to 2016 were collected, and the SARIMA model was established based on the data from January 2004 to June 2016. The SARIMA model fitted and actual values and time factors for the monthly incidence of mumps in 2004-2015 were used to train the SARIMA-GRNN combination model. The data from January to June in 2016 were used for verification, and the model optimal smoothing factor (spread) was screened. The data about the period from July to December in 2016 were introduced back to the established models to evaluate their predictive capacity. The indexes of fitting efficiency and predictive capacity of the models includedmean absolute error percentage (MAPE), mean error rate (MER), mean square error (MSE), and mean absolute error (MAE). Results The SARIMA (0,0,2)(0,1,1)12 model was the most appropriate SARIMA model. The spread value of SARIMA-GRNN combined model was 0.0037. In the fitting phase, the MAPE, MER, MSE and MAE fitted by the single SARIMA model were 16.19%, 15.18%, 0.14% and 0.25%, respectively, and those fitted by the SARIMA-GRNN combination model were 2.93%, 2.28% 0.01% and 0.04%, respectively. The MAPE, MER, MSE and MAE predicted by the single SARIMA model were 17.40%, 17.26%, 0.03 and 0.16, respectively, and those predicted by the SARIMA-GRNN combination model were 15.24%, 15.50%, 0.02 and 0.14, respectively. Conclusions The SARIMA-GRNN combination model is superior to the SARIMA model in fitting and prediction of monthly incidence of mumps, but the prediction accuracy needs to be further improved.
刘天, 姚梦雷, 黄继贵, 吴杨, 陈琦, 童叶青, 陈红缨, 梅芳盛. SARIMA-GRNN组合模型和SARIMA模型在流行性腮腺炎发病率预测中的应用[J]. 实用预防医学, 2021, 28(1): 108-112.
LIU Tian, YAO Meng-lei, HUANG Ji-gui, WU Yang, CHEN Qi, TONG Ye-qing, CHEN Hong-ying, MEI Fang-sheng. Application of SARIMA-GRNN combination model and SARIMA model to predicting the incidence rate of mumps. , 2021, 28(1): 108-112.
[1] Yung CF, Ramsay M. Estimating true hospital morbidity of complications associated with mumps outbreak, England, 2004/05[J]. Eur Surveill, 2016, 21(33):30320. [2] Ho YC, Su BH, Su HJ, et al. The association between the incidence of mumps and meteorological parameters in Taiwan [J]. Hum Vaccin Immunother, 2015, 11(6):1406-1412. [3] Cui A, Zhu Z, Hu Y, et al. Mumps epidemiology and mumps virus genotypes circulating in mainland China during 2013-2015[J]. PLoS One, 2017, 12(1):e0169561. [4] Yung CF,Andrews N,Bukasa A,et al. Mumps complications and effects of mumps vaccination, England and Wales, 2002-2006[J]. Emerg Infect Dis, 2011, 17(4):661-667. [5] Xu Q, Li R, Liu Y, et al. Forecasting the incidence of mumps in Zibo City based on a SARIMA model[J]. Int J Environ Res Public Health, 2017, 14(8):925. [6] 柳林,刘继锋.西安市2007—2016年流行性腮腺炎流行病学分析及发病趋势预测[J].国外医学(医学地理分册),2017,38(3):222-224,227. [7] 王燕.应用时间序列分析[M].第3版.北京:中国人民大学出版社, 2012:141-147. [8] 陈明.MATLAB神经网络原理与实例精解[M]. 清华大学出版社:2013,268-269. [9] 严薇荣.传染病预警指标体系及三种预测模型的研究[D]. 武汉:华中科技大学, 2008. [10] 王彤.医学统计学与SPSS软件应用[M]. 北京大学医学出版社, 2008:338-340. [11] 王雅文,沈忠周,严宝湖,等.ARIMA模型和ARIMA-GRNN模型在AIDS发病预测中的应用[J].中华疾病控制杂志,2018,22(12):1287-1290. [12] 刘天, 张丽杰, 翁熹君, 等. SARIMA模型参数设置探讨[J]. 实用预防医学, 2019, 26(12):1530-1533. [13] Wei W, Jiang J, Liang H, et al. Application of a combined model with autoregressive integrated moving average (ARIMA) and generalized regression neural network (GRNN) in forecasting hepatitis incidence in Heng County, China[J]. PLoS One, 2016, 11(6):e0156768. [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. [15] 马洁,田野,黄璐,等. ARIMA乘积季节模型与广义回归神经网络模型在布鲁菌病发病预测的比较[J]. 山东大学学报(医学版),2018,310(6):76-82. [16] 黄璐,孙娜,许小珊,等.广义回归神经网络在布鲁氏菌病预测中的应用[J].郑州大学学报(医学版),2018,53(6):751-754. [17] Zhang GL, Huang SQ, Duan QH, et al. Application of a hybrid model for predicting the incidence of tuberculosis in Hubei, China[J]. PLoS One, 2013, 8(11):e80969. [18] Ren H,Li J,Yuan ZA,et al. The development of a combined mathematical model to forecast the incidence of hepatitis E in Shanghai, China[J]. BMC Infect Dis, 2013, 13(1):421-421. [19] Zhang X, Liu Y, Yang M, et al. Comparative study of four time series methods in forecasting typhoid fever incidence in China[J]. PLoS One, 2013, 8(5):e63116.