Risk customer identification method and device based on multi-model fusion, equipment and storage medium
By employing a multi-model fusion approach to identify risky customers, combining semantic recognition and discrimination models, the problems of high false alarm rates and low review efficiency in existing technologies have been solved, enabling accurate identification and efficient review of risky customers.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- CHINA MERCHANTS BANK
- Filing Date
- 2026-06-02
- Publication Date
- 2026-07-17
AI Technical Summary
Existing list filtering systems suffer from high false positive rates, high rule maintenance costs, difficulty in enumerating all risk scenarios, and a lack of contextual semantic understanding capabilities when identifying high-risk customers, resulting in low review efficiency.
A multi-model fusion approach is adopted, which processes cross-border transaction message data through semantic recognition and discriminant models, combines text classification and large language models for hierarchical and segmented semantic recognition, and uses a gradient boosting machine model for multi-dimensional feature fusion judgment to achieve accurate identification of risky customers.
It significantly reduced the false alarm rate, improved review efficiency and identification accuracy, achieved risk customer classification and triage and contextual semantic understanding, and reduced the amount of manual review.
Smart Images
Figure CN122415209A_ABST