Aimed at the problem that the multifunctional radar behavior state is complicated and difficult to identify, a multifunctional radar behavior data set is constructed, and a method of radar behavior identification based on neural network is proposed. Firstly, the data are preprocessed to extract the parameter characteristics and behavior state characteristics of the multifunction radar from and to establish a mapping relationship between them. Then, the original radar signal pulse sequence is segmented by the Bayesian rule based on the change point detection algorithm, and the missing characteristic parameters are supplemented to construct a complete pulse array sample to train. Finally, the database is enriched by data reasoning, providing reliable data preparation for datadriven intelligent identification method and enhancing the generalization ability of neural network. BP neural network is designed to train and test the characteristics of the processed radar behavior data set. The simulation results show that the trained network model overcomes the influence of noise variables and other disturbances to some extent, and the accuracy can reach 89%.