Opinion mining and machine learning analysis : What emotions twitter data tell us about telemedicine?

Abstract

The Covid-19 pandemic has made telemedicine one of the most relevant topics in recent years, so data mining about telemedicine obtained from Twitter offers a unique opportunity. The motivation of this study is to identify the emotional groupings of Twitter data users’ opinions for telemedicine using opinion mining techniques such as Sentiment Analysis combined with high-dimensional data classification and prediction methods. Data was collected from Twitter in August and September of 2021 related to telemedicine and official World Health Organization and World Bank documents for the years 2018 and 2021, respectively. Sentiment Analysis showed that 56.2% (n = 5351 tweets) of the sample had predominantly positive emotions toward telemedicine. Telemedicine, telehealth, and ivermectin are the most frequent words in the word cloud. Twitter data users’ opinions consist of nine mood classes (SIL index = 0.80) and the differences between these classes are statistically significant in terms of positive (p < 0.05), negative (p < 0.05), and neutral (p < 0.05) emotions. Neural Network (AUC = 0.842, F1 = 0.492) and Random Forest (AUC = 0.841, F1 = 0.494) are the best predictors of Twitter users’ mood classes compared with other machine learning techniques. The pythagorean tree generated by Random Forest showed that retweet is the best predictor of Twitter users’ opinions and emotional classes towards telemedicine. Future studies will create big social media datasets for a deep understanding of the emotional classes of individuals towards telemedicine technologies.