COVID-19 Outbreak Forecasting Based on Vaccine Rates and Tweets Classification.

Area of Science:

• Computational epidemiology
• Public health informatics
• Machine learning applications

Background:

• COVID-19 spread necessitates advanced surveillance systems for effective containment.
• Existing methods struggle with real-time tracking and early outbreak detection.
• Social networking sites, like Twitter, offer valuable data for disease monitoring.

Purpose of the Study:

• To develop and evaluate a novel machine learning framework for early and accurate COVID-19 detection.
• To leverage social media data and vaccination rates for improved outbreak prediction.
• To provide healthcare decision-makers with a reliable tool for virus containment.

Main Methods:

• A hybrid framework combining word embedding techniques (TF-IDF, FastText, GloVe) for tweet classification.
• Machine learning models (SVM, RF, LR) for classification and (LSTM, Prophet, SVR) for daily prediction.
• Integration of external features including vaccine rates and confirmed cases for enhanced forecasting.

Main Results:

• The proposed framework demonstrated high accuracy, exceeding 80% in COVID-19 detection and prediction.
• Hybrid word embedding methods significantly improved tweet classification performance.
• Incorporating vaccine data alongside social media and case data enhanced prediction reliability.

Conclusions:

• The developed machine learning framework offers a reliable and accurate method for early COVID-19 detection and surveillance.
• Utilizing Twitter data and vaccine rates can significantly improve public health monitoring systems.
• This approach provides a valuable early warning system for health officials to manage outbreaks.