earticle

영어

Considering the shortcomings of the conventional BP neural network, such as slow learning speed, weak anti-interference ability and easy to fall into local minimum, the detection accuracy of P2P traffic detection model is low and the speed is slow, the particle swarm optimization algorithm is used to optimize it here. As the conventional algorithm's optimization ability is the initial parameters, the algorithm is easy to be early, and the convergence speed is slow. Therefore, grouping, organizing, fission and mutation operation on the conventional algorithm have been carried on in order to improve the defect of conventional algorithm. Finally, the P2P traffic detection model is built by using MATLAB software, and traffic detection experiments are carried out on Bittorrent, EMule, PPlive and PPStream 4 P2P network applications. The test data show that the average recognition rate of the recognition model is 96.14%, which is 13.3% higher than that of the conventional PSO-BP model, and9.4% higher than that of the QPSO-BP recognition model for the four P2P network applications.