摘要
目的探讨季节自回归滑动平均模型(seasonal autoregressive integrated moving average,SARIMA)、支持向量回归模型(support vector regression,SVR)及SARIMA-SVR组合模型在全国痢疾发病数预测中的应用效果,为痢疾的防控工作提供科学依据。方法采用2010—2021年我国痢疾月发病数作为训练集,应用Python软件构建SARIMA模型、SVR模型和SARIMA-SVR组合模型,预测2022年1至8月痢疾月发病数趋势,并将其与实际资料进行对比,采用均方根误差(root mean squares error,RMSE)和平均绝对误差(mean absolute error,MAE)衡量模型间的预测效果。结果2010—2021年各年份痢疾发病率的差异有统计学意义(χ^(2)=30.747,P<0.001),总体呈下降趋势(χ^(2)趋势=30.639,P趋势<0.001)。全国痢疾发病趋势具有季节性,其中发病数前3顺位的分别为7月(246755例,14.58%)、8月(244026例,14.42%)和6月(209647例,12.39%)。3种模型中SARIMA-SVR组合模型预测效果的评价指标RMSE、MAE值分别为384.44和282.88;SVR模型RMSE、MAE值分别为436.88和336.71;SARIMA模型RMSE、MAE值分别为704.10和583.10。结论SARIMA-SVR组合模型对我国痢疾发病的预测精度高于单一预测模型,能较好地预测我国痢疾发病趋势,可以用于全国痢疾的中短期预测。
Objective To explore the effectiveness of Seasonal Autoregressive Sliding Average Model(SARIMA),support vector regression(SVR)model and SARIMA-SVR combined model in predicting the incidence of dysentery in China,and to provide scientific basis for the prevention and control of dysentery.Methods Based on the Python software,we used the monthly incidence of dysentery in China from 2010 to 2021 as the training set,and developed SARIMA model,SVR model and combined SARIMA-SVR model to predict the trend of monthly incidence of dysentery from January to August 2022,and compared them with the actual values.Results The incidence of dysentery in each year from 2010 to 2021 was statistically significant(χ^(2)=30.747,P<0.001).The overall trend was downward(χ^(2) trend=30.639,Ptrend<0.001).The incidence of dysentery in China was seasonal,and the top 3 cases were in July(246755 cases,14.58%),August(244026 cases,14.42%)、June(209647 cases,12.39%).Among the three models,the SARIMA-SVR combined model had the best prediction effect for all evaluation indicators,with RMSE and MAE values of 384.44 and 282.88 respectively;Followed by the SVR model,with RMSE and MAE values of 436.88 and 336.71,respectively;The RMSE and MAE values of SARIMA model were 704.10 and 583.10 respectively.Conclusions The combined SARIMA-SVR model is more accurate than a single model in predicting national dysentery incidence,and it can predict the incidence trend of dysentery in our country well and can be used to predict the short and medium term dysentery in the whole country.
作者
王婷
贺湘焱
WANG Ting;HE Xiangyan(School of Public Health,Xinjiang Medical University,Urumqi 830054,Xinjiang Uyghur Autonomous Region,China.;People’s Hospital of Xinjiang Uygur Autonomous Region,Urumqi 830001,Xinjiang Uyghur Autonomous Region,China.)
出处
《预防医学情报杂志》
CAS
2023年第11期1345-1351,共7页
Journal of Preventive Medicine Information
关键词
季节自回归移动平均模型
支持向量回归模型
痢疾
发病预测
seasonal autoregressive moving average model
support vector regression model
dysentery
incidence prediction