The invention provides a classification method based on a hyper-graph transformation network, which is used for solving the problem that in the prior art, high-order semantic information in a heterogeneous network cannot be deeply explored, so that classification cannot be accurately carried out. According to the method, an end-to-end hyper-graph transform network (HGTN) is provided, the communication ability between hyper-edge amplification nodes is used for learning a high-order relation, and semantic information between different types of nodes is mined. Specifically, an attention mechanism is utilized to distribute weights for different types of hyper-graphs, high-order semantic information implied in an original heterogeneous hyper-graph is subjected to cascade learning, useful meta-paths are generated, node embedding features are learned in an end-to-end mode, and a node classification task is completed. The method has good accuracy and universality, and is suitable for node classification tasks of heterogeneous networks such as citation networks, media networks, social networks and the like.