Efficient estimation of influenza epidemics based on seasonal ARIMA models

Abstract

The issue of big data has received a lot of attention from researchers working in various fields as big data sets have been considered to have great potential in detecting or predicting future events such as epidemic outbreaks and changes in stock prices. Reflecting the current popularity of big data, a number of studies have been published to propose models that track influenza epidemics by using internet-based information. The most recent version of influenza tracking model was ARGO, which harnesses search queries from Google and reports from the Centers for Disease Control (CDC). Although the ARGO appears to predict the outbreaks of influenza accurately and outperform other existing methods such as Google Flu Trends (GFT), in this study, a classical seasonal autoregressive and integrated moving average (SARIMA) model is demonstrated to show that it outperforms the ARGO. Compared to the ARGO, the time series model incorporates more accurate seasonality of the past influenza activities, and takes less input variables into account since it utilizes only the reports provided by the CDC. The findings in this study show that the SARIMA model is a more efficient tool for estimating influenza epidemics.