Framework for Detecting Fake Retweets Using Deep Neural Network

• Published in: 2022 IEEE 7th International conference for Convergence in Technology (I2CT) pp. 1 - 6
• Main Authors: Dinesh Hegde, Sampad, Shetty, Akhilesh, Manoj, NM, Kalasad, Abhigna, Bharathi, R
• Format: Conference Proceeding
• Language: English
• Published: IEEE 07.04.2022
• Subjects: BI-LSTM; Deep learning; Deep Neural Network; Entertainment industry; Fully Connected Neural Network; LDA; Natural language processing; Neural networks; Social networking; Social networking (online); Support vector machines; tweets
• DOI: 10.1109/I2CT54291.2022.9824364

Social networking has been increasingly popular in recent years. For news, chatting, entertainment, and other reasons, the majority of people are linked to social media. On social media, misinformation, false data, and conspiracy theories might rapidly spread. A "trending subject" or simply a "trend" on Twitter, for example, is a term, phrase, or topic that receives more attention than others. Retweets help to increase the number of people who see a tweet. Fresh tweets with hashtags related to a trending topic are added to a list of trending tweets to boost the trend's popularity. A classifier employing topic modelling and deep neural networks is established to determine whether these retweets are genuine or related. Deep learning techniques have recently been observed to be useful in the field of natural language processing. Two different frameworks are established namely LDA and Fully Connected Neural Network (FCNN-LDA), and the other one is bag of words and Bidirectional- Long Short Term Memory (Bi-LSTM-BoW). On the validation dataset, the hybrid classifier framework combining both LDA and Fully Connected Neural Network (FCNN-LDA)produced the best results, with 91.8 percent accuracy. The results reveal that the proposed classification system effectively distinguishes tweets and retweets as fake or authentic.