Mechanical deployable reentry vehicles can achieve aerocapture and decelerate effectively because of their large aerodynamic areas. The design and optimization of aerodynamic profile parameters directly affect the deceleration effect. Aiming at the large amount of computation and time consumption of computational fluid dynamics （CFD） for reentry vehicle shape optimization， an approximate calculation optimization method based on the back propagation （BP） neural network is proposed. First， based on the parametric modeling of the reentry vehicle， the orthogonal design is used to generate the samples. Second， the high-precision aerodynamic performance is calculated by CFD. Variance analysis is carried out on the results of sample calculation. Third， the nonlinear fitting for the sample set is conducted by the BP neural network， and the approximate aerodynamic performance model is built. Finally， the multi-island genetic algorithm and the BP neural network model are used to optimize the aerodynamic shape design with the greatest resistance， and the parameter sensitivity analysis is carried out for the optimization results. The results show that the optimization method can quickly and accurately solve the optimization model. The proposed approach ensures the accuracy and improves the computational efficiency， thereby provides a reference for future engineering design and application.