A neural netwok based approach to detect influenza epidemics using search engine query data

Influenza epidemics detection is critically important in recent years because there is a significant economic and public health impact associated with the influenza epidemic. Influenza epidemics detection attracts much attention from governments, organizations, and research institutes, and recently, a novel method using search engine query data to detect influenza activities was presented by Google. In this paper, a data mining based framework using web data is introduced for influenza epidemics detection. Under the framework, a neural network based approach using search engine query data is developed to detect influenza activities. In the proposed method, an automated feature selection model is firstly constructed to reduce the dimension of the query data. Secondly, various neural networks are employed to model the relationship between influenza-like illness data and query data. Thirdly, an optimal neural network is selected as the detector by using the cross-validation method. Finally, the selective neural network detector with the best feature subset is used to detect influenza activities. Experimental results show that the proposed method can outperform traditional statistical models and other models used in the experiments in terms of accuracy. These findings imply that data mining, such as neural network method, can be used as a promising tool to detect influenza activities.