Graph clustering method based on motif structure enhancement

The invention discloses a graph clustering method based on motif structure enhancement. Graph data not only contains feature information of the data, but also has a relation between the data, namely structural information. The invention aims to solve the problems that in the prior art, mining pattern information of graph data is insufficient in the aspect of clustering of the graph data, and a neural network is not fully utilized to deeply learn representation of the graph data. The method comprises the steps that 1, a graph enhancement module enhances the structure of the graph data based on a motif to obtain an enhanced adjacency matrix; 2, inputting the enhanced adjacent matrix and the original feature matrix of the graph data into a depth representation learning module to learn to obtain effective representation of the graph data; 3, unifying representations learned by the depth auto-encoder and the graph auto-encoder by using a self-supervision module; and 4, performing clustering prediction by using the