Quantum behaved binary gravitational search algorithm with random forest for twitter spammer detection

Abstract

• The research stresses the need of using exact and relevant parameters to characterise Twitter spammers. Textual content, URL linkages, user actions, user profiles, and automation are vital for spam identification. • The research uses Random Forest (RF), Binary Gravitational Search Random Forest (BGSRF), and the suggested Quantum Behaved Binary Gravitational Search Random Forest to test the importance of different feature types. QBGSRF beats others, especially in combined feature areas. • Textual characteristics had the greatest F-measure scores across datasets for Twitter spammer detection. This shows that tweet content analysis can detect spam trends. • The study finds user profile information to be another important spam detection factor. It seems that user attributes may greatly improve spam categorisation. • The report finishes by summarising significant results and suggesting future research, highlighting the need for spam detection improvements and the potential for machine learning and quantum algorithms. The emergence of social media platforms like Twitter has significantly changed the landscape of communication by increasing accessibility for widely disseminating official announcements, professional interactions, and important news in real-time. Despite these advantages, the prevalence of spammers and their spamming activities is increasing regularly. To mitigate the growing number of spammers, it is essential to develop an efficient and robust method for Twitter spammer detection. This research presents a novel QBGSRF method by combining the quantum-behaved binary gravitational search algorithm (QBGSA) with random forest (RF) for timely detection of Twitter spammers. The QBGSA algorithm adds the characteristics of quantum computing (QC) and binary gravitational search algorithm (BGSA), which enables the quantum agents to quickly determine solutions using the superposition attributes of QC and the position update via bit-flipping based on velocity probabilities of the BGSA algorithm. In the proposed QBGSRF method, the quantum agents utilize the aforementioned attributes and the principles of the RF algorithm to construct the decision trees for effectively detecting Twitter spammers. The proposed method is assessed for the datasets of 1KS-10KN and Social Honeypot. In order to access the efficacy of the proposed method, the results are also evaluated using the BGSRF method (a combination of BGSA and RF algorithm) and RF algorithm. The experimental evaluations indicate that the proposed method outperforms the aforementioned and state-of-the-art methods.