Sparse Code Multiple Access (SCMA) is a novel non-orthogonal mutiple access scheme, which maps information on different resources into mutidimensional codewords to improve system mapping diversity. However, the performance of the same codebook may vary greatly between different channels, so we propose a new SCMA codebook optimization design for the difference of the codebook, It can be used in the codebook generation process of multiple channels, and optimize the minimum Euclidean distance between code words by genetic algorithm to achieve the effect of improving the codebook performance. In this paper, based on the codebook design of Golden Angle Modulation (GAM), we optimize GAM with the help of genetic algorithm. The simulation results show that in the process of SCMA system simulation using the codebook generated by the optimized parameters in Gaussian channel and Rayleigh fading channel, the BER performance of the codebook is significantly improved.