基于ARIMA与LSTM模型的乌鲁木齐市百日咳发病预测研究  

Research on pertussis incidence prediction in Urumqi based on ARIMA and LSTM models

作　　者：萧楚瑶 黎婷婷 付若楠 尹钰 邹莹 王培生[2] XIAO Chu-yao;LI Ting-ting;FU Ruo-nan;YIN Yu;ZOU Ying;WANG Pei-sheng(School of Public Health,Xinjiang Medical University,Urumqi,Xinjiang 830011,China;不详)

机构地区：[1]新疆医科大学公共卫生学院,新疆乌鲁木齐830011 [2]乌鲁木齐市疾病预防控制中心,新疆乌鲁木齐830026

出　　处：《现代预防医学》2024年第21期3877-3882,共6页Modern Preventive Medicine

基　　金：乌鲁木齐市卫生健康委员会科技计划项目(202346)。

摘　　要：目的 分析ARIMA模型和LSTM模型在乌鲁木齐市百日咳发病预测中的应用,为百日咳的流行趋势研判提供依据。方法 采用乌鲁木齐市2011—2021年百日咳月报告发病数据建立ARIMA模型和LSTM模型,以2022—2023年的发病数据验证两种模型的预测表现,使用均方根误差(RMSE)和平均绝对误差(MAE)进行模型的预测性能评估,并预测2024年百日咳发病情况。结果 乌鲁木齐市2011—2023年百日咳发病呈上升趋势,存在季节性变化。同时自2023年8月开始百日咳进入高发状态。ARIMA模型和LSTM模型的拟合效果良好,但均对2023年7—12月的预测存在一定差异。LSTM模型(RMSE=32.34,MAE=11.41)的总体预测效果优于ARIMA模型(RMSE=42.81,MAE=14.34)。应用验证效果更好的LSTM模型预测2024年百日咳发病趋势,提示百日咳发病将持续上升。结论 LSTM模型对乌鲁木齐市百日咳发病趋势的预测效果更佳,可为百日咳的监测及疫情防控工作提供借鉴与参考。Objective To analyze the application of the ARIMA and LSTM models in predicting pertussis incidence in Urumqi,providing a basis for assessing the epidemic trend of pertussis.Methods Monthly reported incidence data of pertussis in Urumqi from 2011 to 2021 were used to establish ARIMA and LSTM models.The incidence data from 2022 to 2023 were utilized to validate the predictive performance of the two models.The models ' performance was evaluated using Root Mean Square Error(RMSE) and Mean Absolute Error(MAE),and the incidence of pertussis in 2024 was predicted.Results The incidence of pertussis in Urumqi from 2011 to 2023 showed an upward trend with seasonal variations.Additionally,a high incidence state of pertussis began in August 2023.Both the ARIMA and LSTM models demonstrated good fitting,although there were discrepancies in their predictions for July to December 2023.The overall predictive performance of the LSTM model(RMSE=32.34,MAE=11.41) was superior to that of the ARIMA model(RMSE=42.81,MAE=14.34).The LSTM model,which showed better validation results,predicted a continued increase in pertussis incidence for 2024.Conclusion The LSTM model provides a more accurate prediction of the pertussis incidence trend in Urumqi,offering valuable insights for monitoring and controlling the epidemic of pertussis.

关 键 词：百日咳 ARIMA模型 LSTM神经网络模型 预测 

分 类 号：R516.6[医药卫生—内科学]