Predicting Arabic Tweet Popularity by Use of Data and Text Mining Techniques
Rehab Nasser Al-Wehaibi, Muhammad Badruddin Khan · 2014
Among the emerging "social media" or "social networks" applications facilitating communication between individuals, such as YouTube, LinkedIn, and Facebook, Twitter has become one of the most-used of these applications in Arab countries. Twitter works as a microblog, allowing users to write short messages called "tweets". These messages can be spread among Twitter users by "re-tweeting", as the process of re-sending messages in Twitter is known. The re-tweet is therefore a useful and easy way of measuring the level of tweet spread. The aim of this paper is to discover why some tweets spread more than others. To this end, two types of Arabic tweet features were examined: external features (EF) and internal features (IF). Two external features were found to be the best predictors of tweet popularity: tweet category and tweet size. The text content of Arabic tweets, an internal feature, does not reliably predict tweet popularity. The external and internal features used in one classification model were found to be highly predictive of Arabic tweet popularity.