Enterprise risk conduction prediction method and device based on graph features and storage medium

Abstract

The invention relates to an enterprise risk conduction prediction method and device based on graph features and a storage medium, and the method comprises the steps: 1, building enterprise associationgraph data based on bank data, and storing the enterprise association graph data in a graph database; 2, extracting enterprise-enterprise node pairs used as enterprise risk conduction prediction object samples from the graph database; 3, performing column division on the graph mode characteristics of the corresponding data for enterprise-enterprise nodes, and obtaining a risk conduction edge weight by utilizing logistic regression; 4, constructing a characteristic variable of a LightGBM algorithm model for the corresponding topological structure based on the enterprise-enterprise node pairs,and outputting to obtain each unilateral conduction probability result after the model is trained; 5, performing multilateral probability fusion on each unilateral conduction probability result to obtain a risk conduction integration probability result; and 6, performing logistic regression on the risk conduction integration probability result to obtain a final enterprise risk conduction prediction result. The method has the advantages of accurately predicting enterprise risks and the like.

Description

technical field [0001] The present invention relates to the technical field of financial science and technology, in particular to a method, device and storage medium for predicting enterprise risk conduction based on graph features. Background technique [0002] Risk control and early warning are very important parts of bank operations, and operating a bank is, in a sense, operational risk. In an open world, the economic system is directly or indirectly related, leading to the universality of risk transmission. Common situations include (1) equity relationship: the parent company may dispose of the subsidiary’s equity if it has a credit risk. Subsidiaries in normal operation will have a huge impact, and at the same time, subsidiaries will suffer reputation risks; (2) Guarantee relationship: once the guaranteed company defaults, the compensatory obligation will have a direct impact on the cash flow of the guarantee company that was originally operating normally, and the guara...

AI Technical Summary

Problems solved by technology

On the one hand, the number of companies related to enterprises is generally large, especially when predicting a long risk path, it needs to consume huge labor costs; The importance of the relationship between upstream and downstream is not the same. A generalized investigation cannot effectively warn the process of risk transmission in advance, nor can it describe the chain path of risk transmission well; in addition, risk transmission has a time process from latent to outbreak. Visits can't be found very well, and it is easy to be delayed

[0006] At present, the risk transmission between enterprises mostly relies on the manual analysis of the reviewers with expert experience in the above-mentioned scheme 1, which is time-consuming and laborious, and cannot take into account the overall information and discover it within the incubation period.

Inaccurate or missing predictions leading to risk transmission

Images