Flu season is around the corner. However, techniques used to forecast spread of influenza viruses are not able to provide us with real−time data. Newest reports usually represent two-week old information. Fortunately, Google changes everything! Or almost everything. A new study carried out at the University of Warwick shows that it allows us to get reliable estimates of real disease tendencies immediately.
“We conclude that, using adaptive models, Google Flu Trends data can indeed be used to improve real-time influenza monitoring, even when official reports of flu infections are available with only one week’s delay,” researchers at the University of Warwick claim.
Large-scale social science was expensive and slow endeavor until recently. For instance, nationally representative randomized require enormous investments into time and workforce. However, it seems that this issue is at least in part resolved by widespread use of Internet technologies. People using social networking sites and search engines create large datasets which allow us to reveal valuable facts about various social tendencies, such as unemployment, housing prices or spread of diseases.
For example, it was observed recently that there is a relationship between the number of flu cases and rates of search queries which use keywords related to this illness. “On the basis of this observation, they built a monitoring system for ILI which delivered measurements with a delay of only one day, with data accessible via the service Google Flu Trends,” authors of the study published on Royal Society Open Society say.
Tobias Preiss and his colleague Helen Susannah Moat investigated whether this novel scientific tool combined with past statistics can predict influenza epidemics? In order to address this question, they compared statistics provided by Google with influenza estimates reported by official agencies. Scholars discovered that Internet-based sources can considerably improve our ability to forecast future influenza rates.
“We find that when using Google Flu Trends data in combination with historic flu levels, the mean absolute error of in-sample ‘nowcasts’ can be significantly reduced by 14.4%, compared with a baseline model that uses historic data on flu levels only. We further demonstrate that the MAE of out-of-sample nowcasts can also be significantly reduced by between 16.0% and 52.7%, depending on the length of the sliding training interval, ” they report.
Article: PreisT., Moat H.S., 2014, Adaptive nowcasting of influenza outbreaks using Google searches., Royal Society Open Science 1:140095. https://dx.doi.org/10.1098/rsos.140095, source link.