The invention provides a bayesian optimization design method of a metal-dielectric multilayer columnar optical nano antenna, which comprises the following steps: S1, constructing a structure parametercandidate set according to antenna design requirements; s2, determining an optimization objective function as a far-field directivity coefficient, and setting a constraint function at the same time;s3, setting an initial training parameter set and obtaining a corresponding optimization objective function value and a corresponding constraint function value, taking the three sets as a training set, and setting an initial value of iteration times k; s4, training hyper-parameters in the Gaussian model according to the training set; s5, selecting a next candidate parameter; s6, calculating an objective function value and a constraint function value corresponding to the candidate parameter; and S7, judging whether the objective function values corresponding to the candidate parameters meetingthe constraint conditions meet cut-off conditions or not. The method has the advantages that the method is optimized through Bayesian optimization, the calculation time can be effectively shortened, the design efficiency is improved, and the purpose of quickly optimizing the antenna scattering efficiency is achieved.