Study On Identification Of Influencers Based On Trust Relationships In Social Networks

Influential user or influencer is the key to the enterprise in the social media environment for electronic word of mouth marketing. As an important sociological concept, trust is a mechanism for users to simplify decision making as well as to deal with risk. Therefore, the identification of influencers based on trust has increasingly become an important research content in the field of electronic word of mouth marketing. However, the unlimited accessibility, real time and uncertainty of online social networks have brought great challenges to the trust-based identification of influencers. The investigation on approaches to identifying influencers based on trust by analyzing the characteristics of large-scale heterogeneous trust network is of great significance for enterprises, In this way, the enterprises can implement electronic word-of-mouth marketing based on influencers, and then improve their enterprise brand awareness and sale performance.This study tries to study identification of influencers based on trust relationships in the context of social networks. First of all, the domain-aware trust network extraction approach is investigated in the context of large scale heterogeneous trust networks. Second, based on the domain-aware trust network, an approach to identifying domain-aware influencers in terms of their popularity status in the life cycle is proposed by considering the dynamics of trust. Finally, an approach to identifying influencers is investigated by fusing user trust and distrust relationships. The specific research contents and innovations include the following three aspects:A trust network model is established based on the directed multigraph, and a domain-aware trust network extraction approach is then proposed. A directed multigraph is adopted to model the multiple trust relationships among users in a heterogeneous trust network. A domain-aware trust metric is also designed to measure the degree of trust between users in an online social network. With the designed domain-aware trust metric, the trust strength between users can be measured and easy for an end-user to interpret. A domain-aware trust network extraction approach is proposed in accordance with the trust model and domain-aware trust metric. Based on a real-world dataset, some prevailing trust propagation algorithms are applied in the extracted domain-aware trust network, which proves the validity of our domain-aware trust network extraction approach. The experimental results show that most of the trust propagation algorithms are effective in the domain-aware trust network, and have a high accuracy, recall rate and FScore.(2) A social network model with characteristics of domain-dependency and dynamics is established in accordance with the time-varying hypergraph, and then a novel approach is proposed to identify domain-aware influencers with different lifecycles. With the support of social identity theory, we consolidate the research aspects into a multi-stage framework to identify domain-aware influencers. To model the evolving trait of the social network, this study proposes a time-varying hypergraph model reinforced by a new time dimension. An algorithm is conceived to extract a domain-aware user trust network with regard to a time-varying hypergraph. Reinforced by dimensions of trust, domain, and time, a novel approach of product review domain-aware (PRDA) is conceived to identify effective influencers and categorize them into three types, i.e. emerging influencers, holding influencers, and vanishing influencers, in terms of their popularity status in the life cycle. The experimental results from the real world dataset show that the PRDA approach outperforms both the social-network-based influence evaluating (SNIE) approach and the "popular author" approach.(3) An approach to identifying influencers in the context of mix trust network is proposed by fusing user trust and distrust relationships. The structural properties of mix trust network is investigated according to the degree distribution, correlation coefficient and mix pattern. Based on the PageRank algorithm, a mix trust pagerank (MTPR) metric is designed to measure the influential power of an influencer in the context of mix trust network. Then an approach to identifying influencer is proposed based on MTPR by fusing trust and distrust relationship. Finally, the effectiveness of the approach to identifying influencers based on MTPR is vailidated in the context of mix trust network using the real online trust network data that include trust and distrust relationships. The findings show that the distrust degree of majority of users is higher than their trust degree, while only a small part of the user’s trust degree is higher than their distrust degree. Additionally, the mix trust has a disassortativity according to the experimental results. The identified influencers according to the MTPR approach could have a relatively higher trust degree while maintain a lower distrust degree.