A hybrid model for Evaluation of the trust of online social network Based on neuro-fuzzy technique and evolutionary algorithms

In  this research, a hybrid machine learning model based on neuro-fuzzy and one metaّّ
heuristic algorithm called genetic algorithm is proposed to identify user trust in online social
networks. The proposed model performs automatic recognition of users' trust and provides
insight into the most influential features during the recognition process. As part of the work, a
priority-based competitive cascade model is proposed for competing rumor propagation and anti-
rumor cascades. This model is presented to calculate the belief-based precedence value by which
the user decides to believe the received rumor or counter-rumor during information
dissemination. The effect of rumor and anti-rumor cascades in online social networks is analyzed
by considering a neighborhood-based diffusion approach. Experimental results were obtained by
considering Facebook, Wiki, and Twitter to validate the PCC model for rumor and counter-rumor
dissemination in social networks and to evaluate the proposed DMOG algorithm. The DMOG
algorithm provided significant improvement by saving about 26% (for a precedence value of 1.0)
of users from being influenced by rumors on Facebook compared to 20% and 13% by existing
algorithms. On Twitter, the DMOG algorithm saved about 25 percent of users from being
affected by rumors, while the other algorithms affected 19 percent and 21 percent. In addition, in
the wiki dataset, the DMOG algorithm uses a smaller number of initial users for anti-rumor
propagation compared to existing algorithms, saving about 4% of users from being affected by
the rumor. Finally, from the experimental results, it has been observed that the proposed
algorithm performs better than the existing algorithms by using parameters such as budget, rumor
priority and time delay in introducing anti-rumor.