Opinion Mining Analytics for Spotting Omicron Fear-Stimuli Using REPTree Classifier and Natural Language Processing

Abstract

Abstract: Data has indisputably proven overtime to have a better idea and with the surge of big data in the era of coronavirus, research initiatives in the field of data mining continues to leverage computational methodologies. Owing to the dreadful nature of the Omicron-variant, a fight or flight dilemma readily pervades college communities far reaching implications on work ethics of academic front-liners. This study therefore aim to gain insights from academia-sourced data to unravel fear-stimulus in college communities. The predictive analytics is carried out on college-based opinion poll. The Valence Aware Dictionary for Sentiment Reasoning algorithm is deployed for emotion analytics while the Latent Dirichlet Allocation (LDA) and Latent Semantic Indexing (LSI) are employed for topic modelling. The REPTree algorithm models the fear-spotting decision tree using 10-fold cross-validation. Experimental results shows a high performance metrics of 94.68% on Recall and Precision as the hand-washing attribute is returned as the most significant variable with highest information gain. Results of topic modelling likewise returns non-clinical precautionary measures as fear stimulus while the Vader sentiment analysis shows a 22.47%, 25.8%, and 51.73% positive, negative, and neutral polarity scores respectively, indicative of the academic front-liners’ pessimism towards effective safety measure compliance with non-clinical regulations. Keywords: COVID-19, Omicron, Sentiment Analysis, Topic Modelling.