基于Python语言的ARIMA模型在江西省食源性疾病发病率预测中的应用  

Application of ARIMA model based on Python language to predict the incidence of foodborne diseases in Jiangxi Province

作　　者：陈丽敏 刘成伟[3,4] 梁新民 张强 周厚德[3,4] 游兴勇 刘道峰[3,4] 彭思露 CHEN Limin;LIU Chengwei;LIANG Xinmin;ZHANG Qiang;ZHOU Houde;YOU Xingyong;LIU Daofeng;PENG Silu(School of Public Health,Nanchang University,Jiangxi Nanchang 330006,China;Jiangxi Provincial Key Laboratory of Preventive Medicine,Nanchang University,Jiangxi Nanchang 330006,China;Jiangxi Provincial Center for Disease Control and Prevention,Jiangxi Nanchang 330029,China;Jiangxi Province Key Laboratory of Diagnosing and Tracing of Foodborne Disease,Jiangxi Nanchang 330029,China)

机构地区：[1]南昌大学公共卫生学院,江西南昌330006 [2]江西省预防医学重点实验室,江西南昌330006 [3]江西省疾病预防控制中心,江西南昌330029 [4]江西省食源性疾病诊断溯源重点实验室,江西南昌330029

出　　处：《中国食品卫生杂志》2023年第3期458-463,共6页Chinese Journal of Food Hygiene

基　　金：江西省重点实验室计划(20171BCD40021);江西省卫计委科研项目(SKJP_220211996,202110115)。

摘　　要：目的评估整合移动平均自回归模型(ARIMA)预测江西省食源性疾病月发病率的可行性。方法应用Python软件构建ARIMA模型,以2021年数据验证及评估ARIMA模型预测效能,并对2022年1~6月江西省食源性疾病月发病率进行短期预测。结果江西省2016—2021年食源性疾病发病率总体呈下降趋势,每年的8月份为发病高峰期;预测最佳模型为ARIMA(1,0,0)(1,0,2)_(12),贝叶斯信息准则(BIC)为96.66,模型残差为白噪声序列(P>0.05)。模型预测发病率与实际发病率流行趋势基本吻合,整体均方根误差(RMSE)为0.656,以2021年数据验证模型预测效果,预测值与实际值平均绝对百分误差(MAPE)为11.25%,表明模型外推效果较好。结论ARIMA(1,0,0)(1,0,2)_(12)模型可用于江西省食源性疾病发病趋势的短期预测。Objective To evaluate the feasibility of the autoregressive moving average model(ARIMA)for predicting the monthly incidence of foodborne diseases in Jiangxi Province.Methods The ARIMA model was constructed by Python software,and the data from January to December in 2021 was used as the validation set to evaluate the prediction performance of the ARIMA model.The short-term prediction of the monthly incidence of foodborne diseases in Jiangxi Province from January to June in 2022 was carried out.Results The incidence of foodborne diseases in Jiangxi Province from 2016 to 2021 generally showed a downward trend,with the peak incidence in August each year.The best prediction model was ARIMA(1,0,0)(1,0,2)_(12).The Bess Information Criterion(BIC)was 96.66,and the model residual was a white noise sequence(P>0.05).The predicted incidence rate of the model was roughly consistent with the actual incidence trend,and the overall root mean square error(RMSE)was 0.656.The efficacy of the model was verified by the data in 2021.The mean absolute percentage error(MAPE)between the predicted value and the actual value was 11.25%.It showed that the model extrapolation effect was better.Conclusion The ARIMA(1,0,0)(1,0,2)_(12) model can be used for short-term prediction of the incidence trend of foodborne diseases in Jiangxi Province.

关 键 词：食源性疾病 ARIMA模型 PYTHON语言 预测 

分 类 号：R155[医药卫生—营养与食品卫生学]