Flu Trend Prediction Based on Massive Data Analysis

Tracking the trend of infectious diseases, such as influenza, supports public health departments making timely and meaningful decisions and greatly help to stabilize the country and to save people's lives. Traditional systems for monitoring epidemics rely on subsequently reporting confirmed cases with at least one-week delay from the actual epidemic peak, however, there might have already been collective concerns revealed by the active social media network messages. Therefore, by real-time information examination, it is possible to detect and track the spread of epidemics in advance to monitor epidemic activities. This paper aims to study the influenza trend by collecting three datasets including the statistics of the Centers of Disease Control in Taiwan from 2010 to 2016, the Google Trends Web search data, and the King Net national medical diagnosis and consultation records. Linear methods are used to analyze the relationships among those three datasets, while establishing a pattern of interrelationships. In this article, we propose a linear forecasting framework, which is able to capture the major peaks during the interval in those years with greater epidemics. It proves that the huge amount of online social behavior information can be an indirect medium to monitor influenza activities. The prediction model based on the framework provides early response to flu pandemic.