Abstract: Community classification, an important character of complex network research, can be applied from the advantage of neural network in pattern recognition.The basic idea is that the similarity relationship is introduced to simulate the classification between community and neural network by the model of back propagation neural network and the club membership network that was studied by zachary as the experimental surrounding.The method is to set up the adjacency matrix on the membership, then to build the back propagation neural network on the lab design and the adjacency matrix.The experimental outcomes are testified feasible by studying the known samples and the test samples.