The invention discloses a semi-supervised hyperspectral data quantitative analysis method based on a generative adversarial network. The method comprises the following steps: (1) acquiring hyperspectral sample data; (2) constructing a sample training set and a prediction set, wherein a label-free prediction set sample is also used for semi-supervised training; (3) constructing a regression networkbased on the generative adversarial network, wherein the regression network comprises a generator network for generating a sample and a discriminator/regressor network having the functions of discriminating the authenticity of the sample and outputting a quantitative analysis value at the same time; (4) constructing a loss function of the generative adversarial network, wherein the loss functioncomprises a discriminator loss function, a regressor loss function and a generator loss function with a sample distribution matching function; (5) training the generative adversarial network; and (6)adopting the trained regressor to carry out prediction of the quantitative analysis value. According to the method in the invention, the generative adversarial network is adopted for generating samples, and a sample distribution matching strategy is used for supplementing the existing label-free sample set, so that the precision of hyperspectral quantitative analysis is improved.