An Efficient Design of Trust Aware Link Prediction Method In Online Social Networks

Abstract

Since the development of Web 3.0, Online Social Networks (OSNs) have now become an essential part of human life. OSNs have been used for several services and activities such as purchasing, communication, news, music, and communicating with the service provider to seek services. For all these services trust between the users in OSNs is one of the principal factors to take decisions. That is why designing a prediction model to seek the trust among two users who are not directly connected is required. newlineThe scope of the study is to provide an efficient and effective approach to predict trust in OSNs. The study is divided into three aspects. First, to study different aspects of the trust factors in OSNs and find an efficient way to extract the OSNs in an optimized way. OSNs are huge and several participants don t impact the trust between two specific users. A mechanism is required to extract the users that are important for the OSNs and the context in which these users are interacting with each other. Such information has a significant influence on the trust prediction between the trustor and trustee in OSNs. In most cases, the trustor and trustee are not directly connected. To extract a relatively small network from the Huge OSNs can lead to an efficient trust prediction approach. The study proposed a HACPSO for the sub-network extraction from the OSNs that further used in the trust prediction approach. newlineThe second objective of the study is to address the trust prediction problem without any contextual information. The study first analysis different trust factors that can affect the trust value. The trust rating value, such as biased rating in trustor and trustee, the cohesion in trust rating values of trustor and trustees, distribution of the trust rating values, and the propagation of the values of trust rating. Based on the above parameters, the study purposes the Trust-Based SVD-Matrix factorization approach to predict the trust prediction system that used both implicit and explicit factors of the trust. The study experimented on a real-world dataset and compared it with the state of the art approaches. To analyze the strategies, the study uses two accuracy matrices MAE and RSME, and the experiment found that the proposed approach has a drastic improvement in trust prediction accuracy as compared to other approaches. newlineThe third aspect of our study to design a context-based trust prediction approach in OSNs. The study analysis of two Contexts. One is social contextual knowledge, and the other is social phycology theory. In social networks, source and target users have an impact on a similar context. Such a cricket player has more trust in another cricketer. By the similarity of context, one can reduce the sparsity in the online social network matrix. This leads the study to high accuracy and a more efficient trust prediction approach. The study proposed a content-based trust prediction method GETTrust method that is based on matrix factorization. The proposed method predicts trust based on a particular context by using context similarity and GE ratio to reduce sparsity and improve accuracy, respectively. The proposed model uses both propagations of trust using context similarity and regularize the matrix factorization through GETValue. The experiment is conducted to compare the proposed model with other state-of-the-art models. This experiment demonstrates that the proposed model predicts more accurate even there is no communication available among source and target user. newline newline