Trendingtags—Classification & Prediction of Hashtag Popularity Using Twitter Features in Machine Learning Approach Proceedings

Abstract

Most of the tweets are associated with unique hash tags. Some hash tags become trending in a very short time. Predicting which hash tags will become trending in the near future is of significant importance for taking proper decisions in news media, marketing and social media advertising. This research work is aimed at predicting the popularity and tagging the hash tags using machine learning algorithms. It categorizes the popularity under five classes namely not popular, marginally popular, popular, very popular and extremely popular using content and contextual features. The content features can be extracted from both hashtag string and content of the tweets, whereas the contextual features can be extracted from the social network graph of the users. The features are evaluated based on the metrics such as micro-F1 score and macro-F1 score. The result shows that contextual features are more effective than content features as it has the highest prediction accuracy of 94.4%.