UCred: fusion of machine learning and deep learning methods for user credibility on social media

• Published in: Social network analysis and mining 2022-12, Vol.12 (1), p.54, Article 54
• Main Authors: Verma, Pawan Kumar, Agrawal, Prateek, Madaan, Vishu, Gupta, Charu
• Format: Article
• Language: English
• Subjects: Accuracy; Applications of Graph Theory and Complex Networks; Automation; Bidirectionality; Classification; Cloning; Computer Science; Computer viruses; Credibility; Data Mining and Knowledge Discovery; Datasets; Deep learning; Economics; Entertainment; Game Theory; Hoaxes; Humanities; Law; Machine learning; Methodology of the Social Sciences; Neural networks; News; Original Article; Propagation; Social and Behav. Sciences; Social media; Social networks; Statistics for Social Sciences; Support vector machines; User profiles
• ISSN: 1869-5450; 1869-5469
• DOI: 10.1007/s13278-022-00880-1

Online Social Network (OSN) is one of the biggest platforms that spread real and fake news. Many OSN users spread malicious data, fake news, and hoaxes using fake or social bot account for business, political and entertainment purposes. These accounts are also used to spread malicious URLs, viruses and malware. This paper proposes UCred (User Credibility) model to classify user accounts as fake or real. This model uses the combined results of RoBERT (Robustly optimized BERT), Bi-LSTM (Bidirectional LSTM) and RF (Random Forest) for the classification of profile. The output generated from all three techniques is fed into the voting classifier to improve the classification accuracy compared to state-of-the-art approaches. The proposed UCred model gives 98.96% accuracy, notably higher than the state-of-the-art model.