Trust Inference in Social Networks by Combination of Neural Network and Genetic Algorithm

The trust inference problem in a social network is defined as anticipating the trust level that a user can have to another user who is not directly connected to him on the trust network. This research aims to propose a method for trust inference in a trust network. Previous research studies are mainly limited to one type of trust network, and they cannot be used for different trust networks with different values of trust. In this research, soft computing and neural network model are used to predict trust values. To train the neural network system, genetic algorithm is exploited. One of the main advantages of the proposed method is that, unlike previous methods, it is not limited to one type of trust network, and it can also be used for trust networks with different values of trust. In the proposed method, at first four proposed features are extracted from the trust network, and afterward, the proposed neural network system is trained using these features as well as the genetic algorithm. The proposed method is implemented on the standard trust network and is compared with other similar methods. Experimental results indicate that the proposed method is able to produce more accurate results in comparison with previous methods.