A retweet prediction of Thai tweets

Abstract

© 2016 IEEE. Predicting whether a tweet will get a large number of retweets, i.e. to become viral, has been an interest in Twitter research. This paper presents the retweet prediction of tweets by Thai users. Sixteen attributes were used for the prediction. Based on experimental results, decision tree classifier performed better than neural network and Naïve Bayes. Its accuracy when using status-related attributes was 84.3% while that when using user-related attributes was as high as 98.1%. Individual attributes were further evaluated by using gain ratio and ReliefF criteria. They ranked the following to be helpful to the prediction: public list memberships, the number of followers, account longivity, day and time of tweeting, and embedded media.