| 本文已被：浏览 2673次 下载 2065次 |
码上扫一扫！ |
| ARIMA模型与GRNN模型对肺结核发病率预测的对比研究 |
| 胡晓媛1,吴娟2,孙庆文3,沙琨4,王玲玲5,李敏1* |
|
|
(1. 第二军医大学海军医学系航海特殊损伤防护教研室, 上海 200433;
2. 成都军区总医院药剂科, 成都 610083;
3. 第二军医大学基础部数理教研室, 上海 200433;
4. 第二军医大学训练部信息化办公室, 上海 200433;
5. 解放军309医院全军结核病研究所, 北京 100091
*通信作者) |
|
| 摘要: |
| 目的 比较自回归移动平均(ARIMA)模型与广义回归神经网络(GRNN)模型对于肺结核发病率的预测性能。 方法 根据我国2004年1月至2012年12月的肺结核逐月发病率数据资料,应用Eviews 7.0.0.1建立ARIMA模型,应用Matlab 7.1的神经网络工具箱建立GRNN模型;选取2013年肺结核逐月发病率数据对两种预测模型进行检验,比较预测结果。 结果 ARIMA模型和GRNN模型的Theil不等系数(TIC)分别是0.034和0.059,说明ARIMA模型对我国2013年肺结核逐月发病率的拟合程度优于GRNN模型,ARIMA模型相对误差绝对值仅为GRNN模型的57.19%。 结论 ARIMA预测模型更适合用于我国肺结核发病率的预测;建议尝试组合模型预测肺结核发病率。 |
| 关键词: 回归移动平均模型 广义回归神经网络模型 肺结核 预测 |
| DOI：10.16781/j.0258-879x.2016.01.0115 |
| 投稿时间：2015-04-28修订日期：2015-05-20 |
| 基金项目:中国博士后科学基金(2013M542491). |
|
| Comparative study on ARIMA model and GRNN model for predicting the incidence of tuberculosis |
| HU Xiao-yuan1,WU Juan2,SUN Qing-wen3,SHA Kun4,WANG Ling-ling5,LI Min1* |
(1. Department of Navigation Special Damage Protection, Faculty of Naval Medicine, Second Military Medical University, Shanghai 200433, China;
2. Department of Pharmacy, General Hospital, PLA Chengdu Military Area Command, Chengdu 610083, Sichuan, China;
3. Department of Mathematics & Physics, College of Basic Sciences, Second Military Medical University, Shanghai 200433, China;
4. Office of Informatization, Division of Training, Second Military Medical University, Shanghai 200433, China;
5. Institute for Tuberculosis Research, No. 309 Hospital of PLA, Beijing 100091, China
*Corresponding author.) |
| Abstract: |
| Objective To compare the performance of ARIMA model and GRNN model for predicting the incidence of tuberculosis. Methods ARIMA model was set up by Eviews 7.0.0.1 and GRNN model was set up by neural network toolbox of Matlab 7.1 based on the monthly tuberculosis incidence data from January 2004 to December 2012 in China. Monthly tuberculosis incidence data in 2013 were subjected to the two models for testing, and the results were compared between the two groups. Results The Theil unequal coefficients (TIC) were 0.034 and 0.059 for ARIMA model and GRNN model, respectively, indicating that ARIMA model was better than GRNN model to fit with the monthly incidence of tuberculosis in 2013. The absolute value of the relative error for ARIMA model was only 57.19% of GRNN model. Conclusion ARIMA prediction model is more suitable for predicting the incidence of tuberculosis in China, and it is suggested a combination of models should be used to predict the incidence of tuberculosis. |
| Key words: autoregressive integrated moving average model generalized regression neural network model tuberculosis prediction |
