Heterogeneous graph representation learning method and device assisted by large language model, and medium

The invention relates to a big language model-assisted heterogeneous graph representation learning method and device and a medium, and the training method comprises the steps: obtaining the sample data of a heterogeneous graph, obtaining the total number of sample nodes, the sample features of the sample nodes and the sample structure codes according to the sample data, obtaining the sample similarity between the target sample node and other sample nodes according to the sample features and the sample structure codes, calculating the adjacent sample value of the target sample node according to the total number of the sample nodes and the adaptive parameters, and obtaining the adjacent sample nodes of the target sample node according to the sample similarity and the adjacent sample value; inputting the sample data of the target sample node, the sample data of the adjacent sample nodes and the classification of the adjacent sample nodes into a large language model to obtain prediction classification and a class