Methods Using Social Media and Search Queries to Predict Infectious Disease Outbreaks

Seo, Dong-Woo; Shin, Soo-Yong

3) Collection of data from search engines The daily or weekly trends of the keywords for web search queries can be obtained from Google Trends [14] or Naver DataLab [15] . The newly renovated Naver DataLab only offers weekly trends, and it requires several steps to download the output. Therefore, we only explain how to use Google Trends. Google Trends offers various data patterns starting from the previous minute's data to monthly data based on the selected time range. For example, Figure 2 shows the weekly trend of searches for "influenza" of Google Trends between September 9, 2007, and September 8, 2012 . Trend data associated with the predefined keywords were retrieved by setting the location parameter to "South Korea" and setting the time parameters. The time parameters were based on previous disease outbreak seasons, for example, "May 2015 to Jun 2015" for MERS in Korea. May 11, 2015 was the symptom onset day of the first laboratory-confirmed patient. The results can be downloaded in a CSV format by clicking the arrow located in upper right in the graph. The downloaded CSV file consists of two columns. The first column represents the week by indicating the starting date of the week and the second column is the trend data. The researcher should collect the necessary trend data for each keyword. Table 2 shows example influenza statistics from ILI reports ( Figure 1 ) with Google Trends data for the following five keywords: "New influenza, " "New flu, " "Fever, " "Tamiflu, " and "Flu" (Figure 2 ). As shown in Table 2 , there is a timedelay pattern between the influenza statistics and Google Trends data. The important fact is that the current trend data obtained from Google Trends or Naver DataLab are available as normalized values, not the absolute numbers of searches. According to Google Trends, "Numbers represent search in-terest relative to the highest point on the chart for the given region and time. A value of 100 is the peak popularity for the term. A value of 50 means that the term is half as popular. Likewise, a score of 0 means the term was less than 1% as popular as the peak. "