Twitter Spam Detection Using Word Vector and Binary Classifier

In recent years, social media communication plays an important role in sharing messages, thoughts, comments and ideas around the world. Among them, Twitter is one of the most popular communication mediums. Twitter provides free services for sharing information's that are limited to 140 characters. Each month nearly 40 million new twitter accounts were created. Since usage of twitter increases rapidly, fake accounts were created for posting the spam messages and links for redirecting to phishing websites. Researchers proposed various machine learning methodologies for filtering the spam to handle and maintain the social network security problems. But still detecting spam tweets, malicious link in real world scenario is still a complex task. In the proposed work, a novel deep learning technique is introduced to address the above mention issues. Word Vector model is used for learning the syntax of each tweet. Later by using proceeding representation data set, binary classifier is developed. For experimental purpose 1-month twitter datasets are collected. Proposed work outperforms when compared with various existing methodologies and also to detect the non-text based as well. The ratio of spam is determined using the precision, recall, and F-measure metrics.