高级检索

江西省流行性腮腺炎发病率4种时间序列模型预测效果比较

Comparison of prediction effects of four time series models on mumps incidence rate in Jiangxi province

  • 摘要:
    目的 比较季节性自回归移动平均模型(SARIMA)、指数平滑模型(ETS)、指数平滑空间状态模型(TBATS)和自回归神经网络模型(NNAR)4种时间序列模型对江西省流行性腮腺炎(流腮)发病率的预测效果,为流腮的预防控制提供参考依据。
    方法 收集中国疾病预防控制信息系统中江西省2010年1月1日 — 2019年12月31日报告的流腮发病数和发病率数据,以其中2010年1月 — 2018年12月的流腮报告发病率作为训练集,应用R 4.1.2软件构建SARIMA、ETS、TBATS和NNAR模型,并通过模型预测2019年1 — 12月的流腮发病率,采用均方根误差(RMSE)、平均绝对误差(MAE)和平均绝对百分比误差(MAPE)比较4种模型的拟合和预测效果。
    结果 江西省2010、2011、2012、2013、2014、2015、2016、2017、2018和2019年流腮报告发病率分别为20.49/10万、32.03/10万、31.89/10万、19.95/10万、12.22/10万、14.10/10万、16.56/10万、16.21/10万、14.29/10万和21.14/10万,2010 — 2019年流腮年均报告发病率为19.84/10万;江西省流腮发病具有明显的季节性,每年4 — 7月为发病主高峰,11月至次年1月为发病次高峰;4种模型拟合值的变化趋势均与实际值一致,除SARIMA模型MAPE(32.01%)较高外,其他3个模型的MAPE均 < 15%,其中NNAR模型的拟合值更贴近实际值,其RMSE、MAE和MAPE均最低,分别为0.20、0.14和8.24%;除SARIMA模型外,其他3种模型的预测值变化趋势均与实际值一致,此3个模型的MAPE均 < 15%,其中TBATS和ETS模型的预测效果最好。
    结论 ETS、TBATS和NNAR模型对江西省流腮发病率的拟合和预测效果较好,预测结果可为该地区流腮的防控提供理论指导。

     

    Abstract:
    Objective Compare the prediction effects of the seasonal autoregressive integrated moving average (SARIMA), exponential smoothing (ETS), trigonometric box-cox transform, ARMA residuals, trend, and seasonal components (TBATS), and neural network autoregressive (NNAR) time series models on mumps incidence rate in Jiangxi province, to provide a reference for mumps prevention and control.
    Methods Mumps case and incidence rate data reported in Jiangxi province from January 1, 2010 to December 31, 2019 were collected from the China Disease Prevention and Control Information System. The reported mumps incidence rates from January 2010 to December 2018 were used as the training set, and SARIMA, ETS, TBATS, and NNAR models were constructed using R 4.1.2 software to predict the mumps incidence rates from January to December 2019. Root mean squared error (RMSE), mean absolute error (MAE), and mean absolute percentage error (MAPE) were used to compare the fitting and prediction effects of the four models.
    Results The reported mumps incidence rates in Jiangxi province in 2010, 2011, 2012, 2013, 2014, 2015, 2016, 2017, 2018, and 2019 were 20.49/100 000, 32.03/100 000, 31.89/100 000, 19.95/100 000, 12.22/100 000, 14.10/100 000, 16.56/100 000, 16.21/100 000, 14.29/100 000, and 21.14/100 000, respectively, with an average annual reported incidence rate of 19.84/100 000 from 2010 to 2019. Mumps incidence in Jiangxi province showed obvious seasonality, with the main peak from April to July and a secondary peak from November to January of the following year. The fitted values of the four models showed consistent trends with the actual values. Except for the high MAPE (32.01%) of the SARIMA model, the MAPEs of the other three models were all < 15%, with the NNAR model having fitted values closer to the actual values and the lowest RMSE, MAE, and MAPE of 0.20, 0.14, and 8.24%, respectively. Except for the SARIMA model, the predicted value trends of the other three models were consistent with the actual values, and their MAPEs were all < 15%, with the TBATS and ETS models having the best prediction effects.
    Conclusion The ETS, TBATS, and NNAR models have good fitting and prediction effects on mumps incidence rates in Jiangxi province, and the prediction results can provide theoretical guidance for mumps prevention and control in the region.

     

/

返回文章
返回