Detection of suspicious accounts on Twitter using Word2Vec and sentiment analysis

Abstract

Twitter constantly attempts to suspend dangerous or suspicious accounts. This strategy has been questionably ineffective as users continuously recreate their accounts. As a result, there are many accounts that are left unchecked and potentially dangerous to the promotion of ideals and attacks. In this paper, we present a classification method based on sentiment analysis and word2vec to detect suspicious accounts. We evaluate our approach in real use case of main concern using data crawled directly from Twitter. Our classifier returns high accuracy in detecting suspicious accounts.