Heterogeneous information network user abnormal behavior detection method and system based on attention

The invention relates to an attention-based heterogeneous information network user abnormal behavior detection method and system, and the method comprises the steps: firstly converting the historical interaction data of a certain period of time of a heterogeneous information network into graph data, and enabling each node of the graph data to represent a component object of the heterogeneous information network; the sides of the graph data reflect the relation between composition objects of the heterogeneous information network; then, a target function of a user abnormal behavior detection model is constructed based on a graph neural network, and the model derives an interlayer propagation formula represented by node attributes through neighbor information of attention aggregation nodes; and finally, performing gradient updating on each node attribute representation until all node attribute representations converge to obtain each node attribute representation. And compressing the attribute representation of e