Twitter Health Surveillance (THS) System

Area of Science:

• Computational epidemiology
• Public health informatics
• Social media analytics

Background:

• Social media platforms like Twitter generate vast amounts of real-time data.
• Harnessing this data for public health surveillance presents significant opportunities and challenges.
• Existing methods may lack the integration needed for comprehensive health monitoring.

Purpose of the Study:

• To introduce the Twitter Health Surveillance (THS) application framework.
• To provide an integrated platform for health officials to collect, classify, and analyze health-related tweets.
• To establish a big data warehouse for advanced public health data analysis.

Main Methods:

• Development of the THS framework utilizing open-source tools.
• Implementation of Data Acquisition, Tweet Classification, and Big Data Warehousing services.
• Creation of a labeled dataset of approximately 12,000 tweets for validation.
• Testing of Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) recurrent neural networks for tweet classification.

Main Results:

• Achieved high performance in tweet classification: 96% precision, 92% recall, and 91% F1 score.
• Demonstrated the effectiveness of LSTM and GRU models on the curated dataset.
• Validated the THS system's capability to process and classify health-related social media data.

Conclusions:

• The THS framework offers a promising solution for real-time public health surveillance using Twitter data.
• The system's high accuracy in tweet classification supports its utility for health officials.
• THS facilitates the creation of valuable big data resources for further epidemiological research.