Advanced trust classification in social networks using a triple generative adversarial network-assisted capsule network enhanced by gannet optimization

Over the past ten years, social networks (SN) have evolved into the primary infrastructure for people's everyday activities. Trust classification in social networks involves evaluating the trustworthiness of users or the information they share. Traditional trust classification methods often rely on explicit features, such as user ratings, reviews, and social ties. These methods utilize rule-based to assign trust scores for users or entities. However, they face challenges in capturing the exact nature of trust. In this manuscript, Advanced Trust Classification in Social Networks using a Triple Generative Adversarial Network-Assisted Capsule Network Enhanced by Gannet Optimization (Trust-TripleGAN-CapsNet-GOA) is proposed. In this manuscript, the feature vector has computed to every social network users pair after the raw data from the Sentiment140 dataset, is analysed. The membership of trust is then ranked according to five-class classifications by incorporating Tanimoto Trust Similarity coefficient. Then, Triple Generative Adversarial Network-Assisted Capsule Network (TripleGAN-CapsNet) is used to categorise the trust values of users. Finally, the weight parameters of TripleGAN-CapsNet is optimized by the Gannet Optimization Algorithm (GOA) to enhance the accuracy of the trust behaviour in SN. The proposed TrustTripleGAN-CapsNet-GOA method attains 22.94%, 32.36% and 21.96% higher accuracy, 30.27%, 19.46% and 12.39%, higher precision when analyzed with the existing models, such as a method for trust mirroring assessment under social networks parameters with fuzzy system (Trust-SNP-FS), deep matrix factorization in social networks for trust-aware recommendation (Trust-DMF-SN) and towards time-aware context-aware deep trust prediction on the online social networks (TACADTrust-SN). The simulation outcomes exhibit that the proposed method attains 99% accuracy.