Abstract: Resonant frequency is an important parameter in the design process of microstrip antenna （MSA）. An improved feedforward artificial neural network （ANN） that is trained by the algorithm of particle swarm optimization （P SO） is proposed in this paper. Structure of the ANN can be adjusted adaptively according to the complexity of mapped problem, which means its generalization is good. The resonant frequency of rectangular MSA is modeled based on the improved ANN, and the result is better than available ones obviously. It shows the ANN is effective. The model based on PSO ANN can speed the design of MSA. Furthermore, it improves the level of CAD of MSA.