Improving Retweet Prediction via Tweet Features

Abstract

Twitter is a popular social media platform used to share information. Retweeting another user's tweet is considered a powerful mechanism for disseminating information. Predicting whether a a tweet will be retweet has gained increasing interest in recent years. In this study, we examined the prediction of retweeting and the influence of tweet features on retweeting, using tweets with the hashtag #BlackHistoryMonth, which was a top trend in the US Twitter agenda in February 2023, as our experimental space. The tweet features tested included both directly obtainable features and five features obtained using these features. We used the Random Forest algorithm to classify retweeting based on all of these tweet features. We identified the impact of categorized tweet features on the importance of retweeting on Twitter. By adding our newly tested features, we improved the tweet features category by some 6% in terms of F-1 score. This study can contribute to a better understanding of social media usage and user behavior, as well as the development of marketing and advertising strategies.