The existing method for influenza surveillance and early warning is based on the confirmed case data in China.Due to the time difference between symptoms and diagnosis of patients,the timeliness of early warning has to be improved.Hence,the problem of data fusion and analysis based influenza epidemic prediction is studied.First,the scope selection method is used to find out 16 Internet search keywords classified by influenza prevention,symp-toms and treatment.Then,the Baidu index data of the selected keywords from the 46th week of 2017 to the 31st week of 2019 is collected.And then the antecedent time sequence relationship of keywords is obtained with cross-correlation analysis.After that,the prediction models of multiple linear regression and BP neural network are estab-lished using the data of 4 weeks and 2 weeks ahead respectively.Finally,the statistical analysis software SPSS and Matlab platform are used for models training.The test results show that the prediction effect using BP neural net-work model is better than using multiple linear regression model,and that the prediction result using the data of 2 weeks ahead is more accurate than using 4 weeks ahead.
关键词
流感预测/百度指数/神经网络/多元线性回归/数据融合
Key words
influenza prediction/baidu index/neural network/multiple linear regression/data fusion
维普
万方数据