Comparing Social media and Google to detect and predict severe epidemics

Area of Science:

• Epidemiology
• Public Health Surveillance
• Digital Epidemiology

Background:

• Internet technologies offer valuable tools for early epidemic detection and prediction.
• Electronic surveillance systems can be developed using online data analysis to supplement traditional monitoring.
• Assessing the feasibility of integrating search engine and social network data into epidemic surveillance is crucial.

Purpose of the Study:

• To evaluate the feasibility of building an epidemic surveillance system using internet data sources.
• To compare the effectiveness of Google search data versus Twitter social network data for influenza monitoring.
• To determine which data source yields superior results for real-time epidemic prediction.

Main Methods:

• Real-time data acquisition from Google and Twitter over a 23-week period in Greece.
• Comparison of collected online data with official European influenza surveillance data.
• Analysis using the ARIMA model for weekly data and a customized approximate model for daily data.

Main Results:

• Influenza was successfully monitored throughout the study period.
• Google data demonstrated a high Pearson correlation (R=0.933) and a Mean Absolute Percentage Error (MAPE) of 21.358%.
• Twitter data showed slightly better performance with R=0.943 and MAPE=18.742%, outperforming Google and the alternative model.

Conclusions:

• Both Google and Twitter data are valuable for real-time influenza surveillance.
• Twitter data offers slightly superior accuracy for predicting influenza trends compared to Google data.
• Internet-based surveillance systems, particularly using social media, show significant potential for public health monitoring.