Mantis shrimp is one of the important economic aquatic products in China, which enriches the heavy metal elements in the environment. There is a significant difference in contents of heavy metal elements among mantis shrimps in different sea areas. And thus that the pollution status of fishing sea areas could be inferred according to the contents of heavy metal elements in the corresponding mantis shrimps. That is to say, the origin sea area of shrimp mantis can be traced by the contents of heavy metal elements in it. The data of the contents of the three kinds of heavy metals in mantis shrimps from the corresponding three sea areas were used as the input to create a BP neural network discriminant analysis model, which can discriminate the origin sea area of the corresponding mantis shrimp samples after the well optimization of the models. The result showed, using the trained neural network, 86 samples among a total of 90 can be discriminated correctly, the accuracy rate is 95.6%, of which, the accuracy rate by using the training sets is 98.1%, the accuracy rate by using the validation set is 94.4%, and the accuracy rate by using the test set is 88.9%. The results showed that the discriminatory analysis model based on BP neural network model can be used to analyses the relationship among the different elements with the nonlinear complex systems, so as to discriminate the samples and trace them effectively.