Semantic enhancement-based anti-fraud image feature derivation method and system

The invention relates to an anti-fraud image feature derivation method and system based on semantic enhancement, and the method comprises the following steps: S1, building a risk control symbol inference rule and a knowledge graph, and configuring a risk control knowledge rule set which comprises a subgraph mode matching rule class and an edge cutting mode matching rule class; s2, the configured knowledge graph and risk control knowledge rule set are imported into the initial graph in a point-edge graph mode, semantic fusion processing is conducted on the initial graph through an encoder to obtain corresponding hidden variables, and then decoding processing is conducted on the hidden variables to obtain a target graph. According to the anti-fraud image feature derivation method and system based on semantic enhancement provided by the invention, semantic fusion is performed on the initial image to obtain the hidden variables, the hidden variables are used as the derivation features to perform anti-fraud rule o