AI-driven self-supervised anomaly detection system for financial transactions
The AI-driven self-supervised anomaly detection system solves the problems of scarce fraud samples and high labeling costs in financial transactions, enabling efficient detection of new fraud patterns and online repair of abnormal data, thereby improving the model's adaptability and the accuracy of anomaly detection.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Filing Date
- 2026-04-30
- Publication Date
- 2026-07-17
AI Technical Summary
Existing technologies in financial transactions suffer from problems such as scarce fraud samples, high labeling costs, poor generalization ability to new fraud patterns, and insufficient online repair capabilities for abnormal data, making it difficult to adapt to the conceptual drift in financial scenarios.
The AI-driven self-supervised anomaly detection system includes a data acquisition and preprocessing module, a self-supervised pre-training module, an anomaly detection scoring and adaptive optimization module, and an anomaly interpretation and repair module. Through multi-task self-supervised learning and incremental self-supervised learning, it generates robust transaction representation vectors and normal behavior prototypes for anomaly detection and repair.
It achieves self-supervised anomaly detection without requiring a large number of labeled samples, improves the ability to generalize detection of new fraud patterns, reduces the balance between false positive rate and recall rate, can quickly adapt to changes in user behavior and environment, and reduces the negative impact of abnormal data on downstream business processes.
Smart Images
Figure CN122413232A_ABST