The invention relates to a scenic spot recommendation method based on a gated recurrent unit neural network. With the method, technical problems such as difficult start, data sparsity and neglect of implicit semantics in gaming trajectories are solved. The method comprises the following steps: step 1, collecting tourism data < uj1, sj2 and vj3 >, preprocessing the tourism data, and generating a tourism sequence T for representing the tourism track according to all tourism data of the jth tourist and a time sequence; step 2, inputting the tourism sequence in the step 1 into a gating circulationunit neural network, modeling tourism data through the gating circulation unit neural network, and establishing a gating circulation unit neural network learning model; step 3, taking the tourism sequence T in the step 1 as a data set, inputting the data set into the gating circulation unit neural network learning model in the step 2, and taking other scenic spots in the same batch as negative examples for training; and step 4, defining a loss function, updating a recommendation list, and finishing scenic spot recommendation, thereby better solving the problem. The method can be applied to scenic spot recommendation.