Evaluating Opinion filtered neural network trust model

This paper applies neural network training, validating and testing techniques in evaluating the performance of Opinion filtered neural network trust model (OFNN). OFNN model is used to filter heterogeneous trust recommendation systems and deceptive trust recommendations in P2P networks. The model is evaluated in terms of accuracy, convergence speed and reliability. Both simulation data and real data are applied in the evaluation. The results show that OFNN model has accuracy of 83.3% at a convergence speed of 17 training iterations on the real data. The model has accuracy of 93.75% with an average convergence speed of 4545 iterations based on the simulated trust systems in a P2P network.