Medical insurance fraud detection method, device and equipment and readable storage medium

The invention provides a medical insurance fraud detection method, device and equipment and a readable storage medium, and the method comprises the steps: constructing a plurality of heterogeneous graphs based on a medical insurance data set, including a diffusion graph, a topological graph and a feature graph semantic graph; a confidence fusion optimizer is adopted to optimize a basic view, and multi-channel semantic graph convolution with shared parameters is introduced to effectively fuse different types of meta-path semantic graphs. An adaptive attention mechanism is introduced to allocate weights for different views, so that information from multiple views is fused more comprehensively, and rich features of medical insurance data are described accurately. Three different loss functions are adopted to carry out end-to-end model training, and the complementary advantages of multi-view data are fully played, so that the robustness and generalization ability of a fraud detection model are enhanced. The metho