Using the Scatter of Opinions to Predict Responses to Tweets

The importance of social media, such as Twitter and Instagram, has increased as a result of COVID-19: with social media, users can easily communicate with other people. Accordingly, social media have produced both new problems and benefits; these have occasionally occurred with upsurges in reactions owing to the huge numbers of social media followers. "Flaming" is a negative aspect of an upsurge: it is a rapid increase in blame, harassment, and insults directed at a person. Conversely, it is possible to use an upsurge to identify popular topics or trends, which is a positive aspect. Predicting the reactions to a social media post is currently very difficult; however, the aim is to prevent the above problems and provide support by forecasting upsurges at an early stage. We observed that such upsurges were affected by the scatter of opinions in the responses. If there are two groups of conflicting opinions, the original social media post tended to receive both sets of opinions. Toward predicting the reactions to a social media post, this study investigate the relationship between the upsurge and scatter of opinions. We quantify the scatter of opinions in the responses, which we term the scatter of replies. We enter that scatter into multiple regression analysis by means of clustering, and we evaluated our reaction prediction method. We found that for most settings, the correlation coefficient between the predicted and actual values exceeded 0.6, which is a moderate level of correlation.