An intelligent early warning and auditing method for enterprise financial fraud based on big data

Abstract

The present invention discloses an intelligent early warning and auditing method for enterprise financial fraud based on big data. In this invention, through a hybrid architecture of rule engine and machine learning, known fraud can be quickly filtered and unknown anomalies can be accurately identified. The rule engine extracts fixed rules based on historical cases, can efficiently mark high-risk samples that trigger the rules, and reduces the amount of data processed by subsequent models; the machine learning module combines gradient boosting trees and isolation forests to capture outliers with feature deviations and unknown fraud patterns, integrates risk scoring balance rules with model results, and improves the comprehensiveness of overall risk identification. By adjusting model parameters or supplementing features based on audit results, continuous iterative optimization is carried out to effectively reduce the false alarm rate, adapt to the constantly changing fraud methods, and based on the constructed ternary relationship graph, use graph algorithms to mine hidden correlations, and enhance the depth and targeting of auditing.