基于深度学习 MSTF 序列模型的常州市手足口病风险预测研究*

陆胜男 ,  沈思鹏 ,  郑献智

中国卫生统计 ›› 2026, Vol. 43 ›› Issue (5) : 655 -659.

PDF (2200KB)
中国卫生统计 ›› 2026, Vol. 43 ›› Issue (5) : 655 -659. DOI: 10.11783/j.issn.1002-3674.2026.05.003
论著

基于深度学习 MSTF 序列模型的常州市手足口病风险预测研究*

作者信息 +

Risk Prediction of HFMD in Changzhou based on Deep Learning MSTF Model

Author information +
文章历史 +
PDF (2252K)

摘要

目的 比较干预模型和多变量时空融合(multi-variable spatio-temporal fusion,MSTF)模型在常州市手足口病(hand,foot and mouth disease,HFMD)发病预测中的应用。方法 数据来源于常州市2014—2024年手足口病监测报告病例数据。其中采用2014—2023 年数据作为训练集或内部验证集,分别建立自回归移动平均(autoregressive integrated moving average,ARIMA)模型的扩展模型干预模型和MSTF 模型,使用2024 年发病数据作为测试集。使用均方根误差(root mean square error,RMSE)和平均绝对百分误差(mean absolute percentage error,MAPE)进行模型评价。结果 常州市手足口病的发病有较为明显的季节性,5—7月和10—12月为手足口病的发病高峰。最优的 ARIMA干预模型和MSTF 神经网络在2024年测试集发病数的RMSE分别为242.15和144.52, MAPE分别为328.63和93.67, MSTF模型的拟合效果优于ARIMA 模型。采用MSTF神经网络预测2024年发病结果与实际发病情况一致性较高。结论 MSTF 模型对常州市手足口病发病趋势的预测效果更佳,可为手足口病的监测及疫情防控工作提供借鉴与参考。

Abstract

Objective To compare the application of the intervention model, and multi-variable spatio-temporal fusion (MSTF) model in predicting the incidence of hand, foot and mouth disease (HFMD) in Changzhou.Methods Extension of the autoregressive integrated moving average(ARIMA) intervention models and MSTF model were trained and validated based on the reported HFMD case data in Changzhou City from 2014 to 2023. The predictive performance of the two models was tested using the incidence data of 2024. The root mean square error (RMSE) and mean absolute percentage error (MAPE) were used to evaluate the predictive performance of the models.Results The incidence of HFMD in Changzhou City exhibited a significant seasonal pattern, with peaks occurring from May to July and October to December. The optimal ARIMA intervention model and MSTF neural network yielded RMSE values of 242. 15 and 144. 52, and MAPE values of 328. 63 and 93. 67 in the testing set of 2024, respectively. The fitting effect of the MSTF model was superior to that of the ARIMA model. The prediction results using the MSTF neural network were largely consistent with the actual incidence. Conclusion The MSTF model demonstrates better predictive performance for the incidence trend of HFMD in Changzhou City and can provide references for the monitoring and prevention of HFMD.

关键词

手足口病 / LSTM模型 / Transformer模型 / ARIMA模型 / 干预模型 / 预测

Key words

Hand, foot, and mouth disease / LSTM model / Transformer model / ARIMA model / Intervention model / Prediction

引用本文

引用格式 ▾
陆胜男,沈思鹏,郑献智. 基于深度学习 MSTF 序列模型的常州市手足口病风险预测研究*[J]. 中国卫生统计, 2026, 43(5): 655-659 DOI:10.11783/j.issn.1002-3674.2026.05.003

登录浏览全文

4963

注册一个新账户 忘记密码

参考文献

[1]

国家卫生健康委员会. 手足口病诊疗指南(2018年版)[J].传染病信息, 2018, 31(3): 193-198.

[2]

Robinson C, Doane F, Rhodes A. Report of an outbreak of febrile illness with pharyngeal lesions and exanthem: Toronto, summer 1957; isolation of group A Coxsackie virus[J].Canadian Medical Association Journal, 1958, 79 (8): 615-621.

[3]

刘瑞, 闫敏敏, 魏嵘, 等. 儿童法定传染病流行特征及发病趋势预测分析[J].中国医院管理, 2023, 43(2): 46-48.

[4]

Zhu P, Ji W, Li D, et al. Current status of hand—foot—and—mouth disease[J].J Biomed Sci, 2023, 30 (1): 15

[5]

赖晓蓥, 钱俊. ARIMA—LSTM—XGBoost 加权组合模型在肺结核发病趋势预测的研究[J].现代预防医学, 2021, 48 (1): 5-9.

[6]

王斌, 殷立军, 刘春兰, 等. ARIMA 模型和指数平滑法对重庆市巫山县肺结核报告发病数的预测效果比较[J].预防医学情报杂志, 2025,41(10): 1289-1298.

[7]

周滢, 李刚, 王超, 等. 应用时间序列干预模型分析新型冠状病毒感染疫情对北京市其他乙类呼吸道传染病的影响[J].首都公共卫生, 2023, 17 (1): 19-24.

[8]

夏雪, 闫恩来, 李喜武. Transformer 在时间序列预测中的应用综述[J].信息技术与信息化, 2024(3): 124-128.

[9]

万丽, 任常军, 赵树青, 等. 应用ARIMA 模型预测石家庄市手足口病发病趋势[J].中国卫生统计, 2020, 37 (1): 111-113.

[10]

汤岩. 时间序列分析的研究与应用[D].哈尔滨:东北农业大学, 2007.

[11]

Box GEP, Tiao GC. Intervention Analysis with Applications to Economic and Environmental Problems[J].Journal of the American Statistical Association, 1975, 70 (349): 70-79.

[12]

倪宗瓒 巫秀美, 姚树祥, 等. 应用ARIMA 模型动态分析高危人群的肺癌发病率[J].数理医药学杂志, 2001(4): 294-296.

[13]

Andayani F, Theng L, Tsun M, et al. Hybrid LSTM—Transformer Model for Emotion Recognition From Speech Audio Files[J].IEEE Access, 2022, 10: 36018-36027.

[14]

Zheng X, Qiao L, Hong H, et al. Predicting the incidence of common intestinal infections diseases in Changzhou, China based on environmental factors and deep learning[J].BMC Public Health, 2025, 26(1): 319.

[15]

许晓萌, 崔世恒, 王亚菲, 等. 长短时记忆神经网络模型在河北省麻疹疫情发病趋势预测中应用[J].中国公共卫生, 2023, 39 (11): 1464-1468.

[16]

Vaswani A, Shazeer N, Parmar N,et al. Attention Is All You Need[C].NeurIPS 2017 proceadings, 2017.

[17]

Zhou P, Shi W, Tian J, et al. In Attention—Based Bidirectional Long Short—Term Memory Networks for Relation Classification[C]//Association for Computational Linguistics. Berlin, 2016: 207-212.

[18]

袁春燕, 潘英姿, 赵学成, 等. ARIMA 模型在常州市武进区手足口病发病趋势中的应用[J].江苏预防医学, 2022, 33 (6): 707-708.

[19]

李明阳, 程晓庆, 魏叶, 等. 2010—2022年江苏省手足口病暴发疫情流行病学特征及其影响因素分析[J].中国公共卫生, 2024, 40 (3): 336-341.

[20]

黄琼姿, 朱瑞, 徐龙发, 等. 致手足口病柯萨奇病毒A6 型和A10型相关疫苗的研究进展[J].病毒学报, 2023, 39 (4): 1193-1202.

[21]

Lei X, MacKenzie CA. Comparing different models to forecast the number of mass shootings in the United States: An application of forecasting rare event time series data[J].PLoS One, 2023, 18 (6): 1932—6203.

[22]

Shafi SM, Chinnappan SK. Hybrid transformer—CNN and LSTM model for lung disease segmentation and classification[J].PeerJ Comput Sci, 2024, 10: 2376-5992.

[23]

张忆汝, 汤永, 朱敏,等. 基于自然语言处理和深度学习的急性呼吸道传染病早期识别模型的构建[J].中华医院感染学杂志, 2024, 34 (15): 2394-2400.

[24]

彭文祥, 张德英. 基于LSTM 与Transformer的地面沉降智能预测方法研究: 以上海市为例[J].时空信息学报, 2024, 31 (1): 94-103.

基金资助

南京医科大学常州公共卫生高等研究院开放课题基金资助(CPHM202301)

国家自然科学基金面上项目(82373685)

AI Summary AI Mindmap
PDF (2200KB)

0

访问

0

被引

详细

导航
相关文章

AI思维导图

/

〈 〉