Recommendation method and system based on hybrid denoising auto-encoder

The invention discloses a recommendation method and system based on a hybrid denoising auto-encoder, and the method comprises the steps: inputting user-item score data into a pre-trained generative adversarial network to complement missing values in the user-item score data, and for each user, sorting the score values of each item from large to small to generate a recommendation list. An auto-encoder is used as a generator of the generative adversarial network, noise addition is performed on original user-item scoring data during training, and the user-item scoring data after noise addition is used for training, so that the robustness of the generative adversarial network is improved, and the hybrid use of the auto-encoder and the generative adversarial network can complement each other. According to the method and the system, the respective advantages are fully exerted, deeper features of score data are learned, the generative adversarial network obtained after training can fully learn differences between no