摘要
目的提高搜索引擎的网络大数据预测传染病疫情趋势的精确性。方法利用回顾性流行病学调查方法,以2013-2018年广东省H7N9亚型禽流感月度与周度新增病例数为研究对象,基于“H7N9”关键词和H7N9亚型禽流感临床症状关键词的百度指数,分波段建立支持向量机预测模型和多元线性回归预测模型开展拟合度分析。结果新增病例数与“H7N9”关键词百度指数的变化趋势可将疫情划分为4个波段;第2、3波疫情的预测值能较好地描述真实病例数的变化趋势;临床症状的关键词与第2、3波段疫情实际病例数有明显的正相关;对第4波段疫情的预测值比较贴近实际发生率,比支持向量机回归具有更高的预测精度。结论利用公众搜索行为与疫情变化呈现的波段特征以及公众对传染病临床症状关键词的查询频率在很大程度上提高了搜索引擎网络大数据H7N9亚型禽流感疫情趋势的预测能力。
To improve accuracy in predicting the epidemic trends of infectious diseases by using network big data from a search engine,we conducted a retrospective epidemiological investigation of the monthly and weekly new H7N9 subtype avian influenza cases in Guangdong Province from 2013 to 2018.We established a support vector machine(SVM)prediction model and multiple linear regression prediction model to perform fitting degree analysis based on the Baidu index of the keywords"H7N9"and the clinical symptoms of H7N9 subtype avian influenza.We found that the number of new cases and the trend in the Baidu index of the keyword"H7N9"could be divided into four groups.The predicted values of the second and third wave epidemics described the trends in the actual number of cases.The keywords of clinical symptoms were positively correlated with the actual number of cases.The predicted value of the fourth wave epidemic was closer to the actual incidence and had higher prediction accuracy than SVM regression.The results showed that the wave characteristics of public search behavior and epidemics changed the public search frequency for keywords for the clinical symptoms of infectious diseases,thus greatly improving the ability of network big data from a search engine to predict the H7N9 subtype avian influenza epidemic trend.
作者
黄泽颖
HUANG Ze-ying(Institute of Food and Nutrition Development,Ministry of Agriculture and Rural Affairs,Beijing 100081,China)
出处
《中国人兽共患病学报》
CAS
CSCD
北大核心
2020年第11期962-968,共7页
Chinese Journal of Zoonoses
基金
国家自然科学基金青年项目(No.71804078)。
关键词
百度指数
传染病
预测
H7N9亚型禽流感
广东省
Baidu index
epidemics
prediction
H7N9 subtype avian influenza
Guangdong Province
