P2P traffic identification based on bayesian regularization BP neural network

摘要

It will increase identification accuracy when P2P traffic identification method based on flow feature is combined with machine learning methods. Recently, the most applied machine learning method is neural networks, but neural networks has insufficient generalization ability, this paper proposes an identification method based on BP neural network that use bayesian regularization to improve its generalization ability. The simulation results show that this method can effectively improve the identification accuracy in practice.