Detecting spam tweets in Twitter using a data stream clustering algorithm

• Published in: 2015 International Congress on Technology, Communication and Knowledge (ICTCK) pp. 347 - 351
• Main Authors: Eshraqi, Nasim, Jalali, Mehrdad, Moattar, Mohammad Hossein
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.11.2015
• Subjects: Classification algorithms; Clustering; Clustering algorithms; Data stream; Feature extraction; Heuristic algorithms; Software engineering; Spam detection; Twitter
• DOI: 10.1109/ICTCK.2015.7582694

Social network sites are becoming great through millions of users and information have been gathered from the users. This information have equal advantages to their friends and spammers. Twitter is one of the most popular social networks that, users can send short textual messages namely tweet. Researches have shown that this network is subject to spammer's invasion more than other social networks and more than six percent of its tweets are spam. So diagnose of the spam tweets is very important. Firstly, in this research, we determine various features for spam detection and then by using a clustering algorithm based on the data stream, we identify spam tweets. previous works in the field of spam tweets were done by classification algorithms. it is the first time that for spam tweets detection, an data stream clustering algorithm is used. Den stream Algorithm can cluster tweets and consider Outliers as spam. Results show, when this algorithm is set properly the amount of accuracy and precision of spam tweets detection will improve and false positive rate will reach to the minimum value in comparison with previous works.