基于百度指数和ARDL模型的广东省清远市2013-2017年手足口病发病数分析  

Analysis and Prediction of HFMD in Qingyuan City Based on Baidu Indexes and ARDL Model From 2013 to 2017

作　　者：杜玉忠[1] 黄业草 何慧萍[1] 范秀红[1] 卢文涛[1] 蔡永铭[2] 蔡巧[2] DU Yuzhong;HUANG Yecao;HE Huiping;FAN Xiuhong;LU Wentao;CAI Yongming;CAI Qiao(Qingyuan Center for Disease Control and Prevention,Qingyuan 511500,Guangdong Province,China;Guangdong Pharmaceutical University,Guangzhou 510006,Guangdong Province,China.)

机构地区：[1]清远市疾病预防控制中心,广东清远511500 [2]广东药科大学,广州510006

出　　处：《预防医学情报杂志》2018年第11期1388-1391,共4页Journal of Preventive Medicine Information

基　　金：清远市市级科研立项:基于百度指数的清远市手足口病发病预测模型的比较研究(项目编号:2016B161)

摘　　要：目的探讨利用手足口病相关关键词百度指数建立清远市手足口病发病预测模型。方法收集《中国疾病预防控制信息系统》清远市范围内2013-01-01/2017-05-31每周手足口病发病数,同时从百度指数官网收集同期预先定义的手足口病相关关键词在清远市范围内的百度指数,选取与手足口病实际发病数相关系数大于0. 5的纳入分析。以手足口病实际发病数为因变量,与各关键词百度指数建立自回归分布滞后模型(ARDL),并用建立的ARDL模型预测未来1～4周手足口病发病数。结果共收集230周数据,手足口病合计发病57 849例,纳入"手足口病初期症状"(r_s=0. 73)、"手足口病"(r_s=0. 81)、"手足口病图片"(r_s=0. 62)、"手足口病症状"(r_s=0. 60)和"手足口病吃什么药"(r_s=0. 53) 5个关键词,建立了ARDL(2,1,0,0,2,1)模型,对未来第2周的发病预测较为准确(MAPE=5. 94,TIC=0. 03)。结论利用百度指数建立的ARDL预测模型能够较为准确地预测手足口病发病情况。Objective To establish a prediction model for hand - foot - mouth disease （HFMD） in Qingyuan City by using Baidu Index of HFMD key words. Methods The weekly incidence data of HFMD from January 1,2013 to May 31, 2017 in Qingyuan City as well as the Baidu Index related to HFMD in the same period in Qingyuan City were collected. Variables with correlation coefficients above 0. 5 between HFMD cases and the corresponding Baidu Indexes were included in the analysis. Taking the actual incidence of HFMD as the dependent variable, an autoregressive distribution lag （ARDL） model was established was used to predict the number of HFMD cases in the next one to four weeks. Results A total of 57 849 cases of HFMD were collected, and the Baidu Indexes of ＂the initial symptoms of HFMD＂ （rs =0.73）, ＂HFMD＂ （rs =0. 81）, ＂HFMD pictures＂ （ rs = O. 62）, ＂symptoms of hand - foot - mouth diseas＂ （ rs = O. 60 ） and ＂what medicine to take for hand - foot -mouth disease＂ were included in the analysis. An ARDL （2, 1, 0, 0, 2, 1） model was constructed and, with it, the number of HFMD cases in the second week could be accurately predicated （MAPE = 5.94, TIC = 0.03）. Conclusion ARDL prediction model established by using Baidu indexes is able to predict the incidence of HFMD accurately.

关 键 词：百度指数 手足口病 自回归分布滞后模型 预测 

分 类 号：R512.5[医药卫生—内科学]