Flu Outbreak Prediction Using Twitter Posts Classification and Linear Regression With Historical Centers for Disease...

Area of Science:

• Computational epidemiology
• Public health surveillance
• Machine learning applications

Background:

• Social networking sites (SNSs) offer vast real-time data for disease outbreak tracking.
• Conventional machine learning methods are insufficient for novel outbreaks with evolving symptoms.

Purpose of the Study:

• To develop an efficient and accurate framework for tracking disease outbreaks using SNS data.
• To provide early warnings for emerging infectious diseases, including novel ones.

Main Methods:

• A 3-module framework: text classification (FastText vs. ML), mapping, and linear regression.
• FastText (FT) classifier evaluated for efficiency and accuracy in classifying flu-related tweets.
• Weekly flu rate predictions generated using mapped tweet data and historical CDC data.

Main Results:

• FastText achieved 89.9% F-measure for flu tweet classification, proving efficient and accurate.
• Linear regression model demonstrated high accuracy (96.29% correlation with CDC data) for weekly flu rate prediction.
• The framework successfully predicted flu trends with high correlation to ground truth data.

Conclusions:

• The proposed FT-based framework enhances accuracy and efficiency in disease surveillance using SNS data.
• The system is effective for tracking and predicting new outbreaks with novel symptoms.
• This approach offers a valuable tool for real-time public health monitoring.