A low-cost adaptive beamforming system is designed to solve the space radiation interference problem of millimeter wave communication in complex electromagnetic environment. Firstly, the communication system model in complex electromagnetic environment and the objective function are established. Then, the support vector machine algorithm in machine learning is used to simplify the objective function to obtain the ideal coding vector. Finally, the gradient pursuit algorithm is used to build sparse reconstruction of the ideal beam vector to achieve low-cost anti-jamming beam. Simulation results show that the proposed beamforming system can effectively suppress the interference and improve the communication quality.