Implementation of Cosine Similarity and Time Interval Entropy Method to Identify Bot Spammer Account on Twitter

Abstract

Twitter is one of the social media that has many users. However, the popularity of twitter is also followed by the large number of spam sent by automated programs called bot spammers. Bot spammer produces tweets that have similar content with previous tweets, and it may post in regular interval since it was automatically posted based on a scheduler system. This research uses HGrid247 tool and supervised classification method to classify bot spammer account and legitimate user account based on tweet similarity rates and regularity of time interval between tweets. Cosine similarity method used to observe tweet similarity rates while time interval is used to observe regular behavior in posting a tweet. Based on the result of performance evaluation, the proposed method can produce accuracy by 90%. This proves that the combination of cosine similarity dan time interval entropy methods can optimize the identification process of bot spammers in twitter.