摘要
目的:分析乌鲁木齐市流感样病例月发病数的变化趋势,建立长短期记忆(LSTM)模型,对流感样病例例数数进行预测,为乌鲁木齐市流感的预防与控制提供科学依据。方法:利用2015年1月-2018年3月乌鲁木齐市的每月气象数据、流感样病例监测数据,采用单变量LSTM模型和多变量LSTM模型对乌鲁木齐市流感样病例例数的时间序列进行预测,使用RMSE和MAE值评价不同方法的预测精度。结果:单变量LSTM模型和多变量LSTM模型的RMSE值分别是66. 17和56. 91;MAE值分别是60. 42和39. 07。与单变量LSTM模型相比,多变量的LSTM模型预测效果较好。结论:本研究所建立的多变量LSTM模型能较好地预测ILI病例数的发病趋势,为流感监测和预防控制提供依据。
[ Objective ] To analyze the change trend of the monthly incidence of influenza-like cases in Urumqi,establish long-short term memory( LSTM) model, and predict the number of influenza-like cases, so as to provide scientific basis for the prevention and control of influenza in Urumqi. [ Methods ] Monthly meteorological data and influenza-like case monitoring data from January 2015 to March 2018 in Urumqi were used to predict the time series of influenza-like cases in Urumqi using univariate LSTM model and multivariate LSTMmodel, RMSE and MAE values were used to evaluate the prediction accuracy of different methods. [ Results ] RMSE values of univariate LSTM model and multivariate LSTM model are 66. 17 and 56. 91 respectively. The MAE values are 60. 42 and 39. 07 respectively.Compared with the single-variable LSTM model, the multivariable LSTM model has a better predictive effect.[ Conclusion ] The multivariate LSTM model established in this study can better predict the incidence trend of ILI cases, providing a basis for influenza surveillance, preventionand control.
作者
龚风云
王凯
GONG Feng-yun;WANG Kai(College of Applied Mathematics,Xinjiang University of Finance and Economis,Urumqi Xinjiang 830012,China;College of Medical Engineering and Technology,Xinjiang Medical University,Urumqi Xinjiang 830011,China)
出处
《科技视界》
2019年第31期20-22,共3页
Science & Technology Vision
基金
国家自然科学基金资助（2015.01-2018.1211461073）
关键词
LSTM
流感样病例
气象因素
预测
LSTM
Influenza-like cases
Meteorological factors
Prediction
