基于图神经网络的金融征信研究

The evaluation of the credit of users who apply for loans by financial institutions is one of the frontier directions in the field of Internet finance.Firstly,based on the historical data of the Internet financial loan network,the network modeling of the loan relationship between users reflects the complex network of loan correlation that integrates the interaction between user nodes and surrounding relationship nodes.Secondly,by introducing a graph neural network model based on the structural characteristic index of node centrality,a personal credit evaluation model with adjacent circle layer information and loan credit information is proposed.Finally,the model is implemented on a historical data set containing 756 100 transaction records,and is compared with the BP neural network algorithm and the RF-Logistic model.The results show that the proposed model has higher evaluation accuracy.