IDENTIFICATION OF SPAMMERS ON TWITTER USING MODIFIED RANDOM FOREST CLASSIFIER

Abstract

One of the most popular online social networks is twitter that users can share short textual data called tweets. Researchers have defined that this network is subjected to spammer’s attack higher than other social networks and more than 6% of its tweets are spam. The detection of spam tweets is essential. Firstly, in this study, we find various features for spam detection by using a clustering algorithm based on the data stream, and we identify spam tweets. The Classification algorithms performed previous works in the field of spam tweets. It is the first time that for spam tweets disclosure, a data stream clustering algorithm applied. Then stream Algorithm can gather tweets and investigate outliers as spam. Results show when this algorithm set accurately, the number of accuracy and precision of spam tweets discovery will improve, and the false-positive rate will achieve the least value in clustering with early work. This paper suggests an inductive-learning system for the detection of Twitter-spammers and applies a Modified Random-Forest (MRF) mechanism to a restricted set of characteristics obtained from traffic. Experimental results show that the proposed method exceeds current approaches to this problem.