AI BASED ANALYSIS OF SOCIAL MEDIA DATA FOR PUBLIC HEALTH DOMAIN

Abstract

Public health fosters community well-being, provides community security, and protects communities from environmental risks. This field contributes to the population’s access to safe and high-quality health-care. The sources of the public health risks can be environmental, social, economic, physical, chemical, political, and biological. The primary mode of solving public health problems is applying preventive measures, unlike the curative measures used in medicine. The Web3.0 era has resulted in an unprecedented amount of user-generated data being made public via social media platforms like Twitter, Facebook, LinkedIn, Instagram, Snapchat, and others. This user-generated data has become a significant source of gathering user-specific information to solve many real-time challenges. Social media and web data have a number of advantages over traditional data sources, including real-time data availability, ease of access, and lower costs. Social media allows asking, and answering, questions we never thought possible. Much research is currently being done in public health domain, with data from many search engines, social media websites, and health-related websites being used to analyze user perceptions understanding & performing required interventions, collecting user feedback for various policies, analyzing user behavior to predict the user-demands, and detecting significant events & spread of false information, etc. The goal of this thesis is to analyse Twitter data connected to public health actions, mentions, and associations in order to establish frameworks that can help overcome the constraints of previous research. Despite the amount of social media data, the use of high-performance supervised machine learning and data mining models is limited due to the lack of labels. As a result, this thesis has presented solutions that either require unsupervised models or a small number of labels.