Topic and sentiment analysis of responses to Muslim clerics’ misinformation correction about COVID-19 vaccine: Comparison of three machine learning models

Abstract

Abstract Purpose The purpose of this research was to use develop a sentiment model using machine learning algorithms for discerning public response about the misinformation correction practices of Muslim clerics on YouTube. Method This study employed three machine learning algorithms, Naïve Bayes, SVM, and a Balanced Random Forest to build a sentiment model that can detect Muslim sentiment about Muslim clerics’ anti-misinformation campaign on YouTube. Overall, 9701 comments were collected. An LDA-based topic model was also employed to understand the most expressed topics in the YouTube comments. Results The confusion matrix and accuracy score assessment revealed that the balanced random forest-based model demonstrated the best performance. Overall, the sentiment analysis discovered that 74 percent of the comments were negative, and 26 percent were positive. An LDA-based topic model also revealed the eight most discussed topics associated with ten keywords in those YouTube comments. Practical implications The sentiment and topic model from this study will particularly help public health professionals and researchers to better understand the nature of vaccine misinformation and hesitancy in the Muslim communities. Social implications This study offers the joint task force of Muslim clerics and medical professionals, and the future misinformation campaigns a sentiment detection model to understand public attitude to such practices on social media. Originality While the impact of misinformation on public sentiment and opinion on social media has been researched extensively, Muslim perspectives on combating misinformation have received less attention. This research is the first to evaluate responses towards Muslim clerics correcting religious vaccine misinformation using machine learning models.