Influenza-Like Illness Patients Forecasting by Fusing Internet Public Opinion

ABSTRACT

Due to rapid change in influenza viruses, a prediction model for outbreaks of influenza-like illnesses helps to find out the spread of the illnesses in real time. In addition to using traditional hydrological and atmospheric data, popular search keywords on Google Trends are used as features in this research. Google Trends are popular keyword searches on the Google search engine. Popular keywords used in discussions of influenza-like symptoms at specific regions within specific periods are used in this research. Public holiday information in Taiwan, the population density, air quality indices, and the numbers of COVID-19 confirmed cases are also used as features in this research. An Ensemble Learning model, combining Random Forest and XGBoost, is used in this research. It can be confirmed from the actual experimental results in this research that the use of the ensemble learning prediction model proposed in this research can accurately predict the trend of influenza-like cases. The evaluation results show that the mean RMSLE of our proposed model is 0.2 in comparison with the actual number of influenza-like cases. © 2022, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.