广州市登革热传播风险的预测研究

The prediction of dengue fever transmission risk in Guangzhou

目的利用互联网数据百度指数对登革热疫情进行预测,为制定登革热防治策略提供科学依据。方法收集广州市2015-2018年登革热每周发病数及本地区相应每周的登革热关键词的百度指数,通过R语言分析百度指数与实际病例数相关性、建立多元线性回归模型,并通过交叉验证和反向测试评估模型效果。结果回归模型预测值与实际值的Pearson相关系数为0.845(P<0.05),Spearman相关系数为0.608(P<0.05),且数据模型也得到较好验证。考虑到历史新发病例数对当前的影响,改进后的预测模型为Y=0.03X1-0.04X2-0.10X3-0.07X5-0.126X7+0.16X8+0.076X10+0.012X12+0.713Xp+18.30(调整R2=0.896 6,AIC=911.49),预测值与实际值更为接近。结论多元线性回归模型可较好地预测广州市登革热疫情发病趋势。

Objective To predict the dengue fever epidemic based on Baidu Index of internet data,and provide scientific basis for the development of dengue fever prevention strategies.Methods Weekly incidence of dengue fever in Guangzhou from 2015 to 2018 and the local corresponding weekly Baidu Index of the keywords of dengue fever were collected.The correlation between Baidu Index and the actual number of cases was analyzed by R language to establish a multiple linear regression model.Then the model was evaluated by cross-validation and reverse test.Results The Pearson correlation coefficient between predictive values and actual values was 0.845(P<0.05)and Spearman correlation coefficient was 0.608(P<0.05).The regression model was also well validated.Considering the impact of the number of past dengue fever cases on the current dengue fever cases,the improved prediction model was Y=0.03X1-0.04X2-0.10X3-0.07X5-0.126X7+0.16X8+0.076X10+0.012X12+0.713Xp+18.30(adjusted R2=0.8966,AIC=911.49),and the predicted value was closer to the actual value.Conclusion The multivariable linear regression model can predict the dengue fever epidemic in Guangzhou.