This study examines the relationship between internet search engine queries (Google Trends) and social media data (Twitter) on COVID-19 and their potential for predicting COVID-19 cases in Canada. Using data from January 1, 2020, to March 31, 2020, the researchers analyzed symptom keywords and COVID-19-related hashtags, applying signal processing techniques to remove noise. Time-lagged cross-correlation analyses and a long short-term memory (LSTM) model were used for forecasting. Results showed strong correlations between symptom searches and COVID-19 incidence, with searches peaking several days earlier. The LSTM model, using Google Trends data, achieved high accuracy in forecasting. While Twitter data showed correlation, it was less effective than Google Trends for prediction. The study concludes that internet search and social media data can serve as early warning signals for real-time COVID-19 surveillance, though challenges remain in model development.

Yang Yang, Shu-Feng Tsao, Mohammad A Basri, Helen H Chen, Zahid A Butt