Based on the analogue of general uncertainty relation in information transmitting processes, by the introduction of multi correlative coefficient R to manifest the complexity of the function and the hidden unit h to manifest the complexity of the network structure, a general uncertainty relation between the learning ability and the generalization ability suited to overfitting of BP neural network was revealed in the modeling of BP neural network. Tests of numerical simulation for 12 kinds of complicated function were carried out to determine the value distribution (1×10 -5 ～5×10 -4 ) of overfitting parameter in the uncertainty relation. Based on the uncertainty relation, the judgement of overfitting in the training process of gived sample sets using BP network was given.