Classifying and fact-checking health-related information about COVID-19 on Twitter/X using machine learning and deep...

Area of Science:

• Computational linguistics
• Health informatics
• Machine learning

Background:

• Health information on social media presents unique challenges for misinformation detection.
• Robust fact-checking models are needed to combat the spread of health misinformation.
• Previous methods require further investigation for effective health content analysis.

Purpose of the Study:

• To identify the most effective approach for detecting and classifying reliable versus misinformation health content on Twitter/X.
• To compare the performance of various machine learning and deep learning models for health misinformation detection.
• To analyze health information related to COVID-19 shared on Twitter/X.

Main Methods:

• Utilized seven distinct machine learning/deep learning models for analysis.
• Collected, processed, and labeled Twitter/X data into "Trustworthy information" and "Misinformation" datasets.
• Employed cosine similarity for class balancing and evaluated models using accuracy, precision, recall, F1-score, ROC curve, and AUC.

Main Results:

• TextConvoNet demonstrated superior performance with average accuracy, precision, recall, and F1 scores above 90%.
• The model achieved high accuracy (85%) and F1 score (89%) for trustworthy information.
• TextConvoNet excelled in identifying misinformation, with accuracy of 94% and F1 score of 91%.

Conclusions:

• TextConvoNet is the most effective model for detecting and classifying trustworthy versus misinformation health content on Twitter/X.
• The study highlights the potential of deep learning models in combating health misinformation.
• Findings provide a robust framework for future health fact-checking systems.