A FRAMEWORK FOR TWEET CLASSIFICATION AND ANALYSIS ON SOCIAL MEDIA PLATFORM USING FEDERATED LEARNING

Abstract

Social media plays a pivotal role in the daily activities of individuals, serving as a medium for the dissemination of events, activities, and information through various forms of posts, including tweets, status updates, and pictures. The source of information is determined by analyzing the impact of a user's association with a particular tweet. In this research paper, we present a framework based on the principles of Federated Learning (FL) to classify and analyze tweets across different social media platforms. The framework incorporates feature mapping and feature indexing techniques to determine the threshold computation value for categorizing tweets as either "positive" or "negative." Importantly, our framework is platform-agnostic and has been rigorously validated using a diverse dataset comprising dynamic trends and social media posts from platforms like X (formerly known as Tweeter), Koo, and Instagram. Our findings demonstrate that the framework achieved an impressive accuracy rate of 93.54% in classifying TP (Trending Topic) posts with respect to the subject matter under consideration.