Ensemble Approach for Detecting COVID-19 Propaganda on Online Social Networks

Abstract

COVID-19 affected the entire world due to the unavailability of the vaccine. Thesocial distancing was a contributing factor that gave rise to the usage of OnlineSocial Networks. It has been seen that people share the information that comes tothem without verifying its source . One of the common forms of information that isdisseminated that have a radical purpose is propaganda. Propaganda is organizedand conscious method of molding conclusions and impacting an individual'scontemplations to accomplish the ideal aim of proselytizer. For this paper, differentpropagandistic tweets were shared in the COVID-19 Era. Data regarding COVID-19propaganda was extracted from Twitter. Labelling of data was performed manuallyusing different propaganda identification techniques and Hybrid feature engineeringwas used to select the essential features. Ensemble machine learning classifiers wereused for performing the binary classification. Adaboost shows an accuracy of98.7%, which learns from a weak learning algorithm by updating the weights.