Agent-based anti-fraud business processing system and method, medium, device

By using an agent-based anti-fraud business processing system, the system preprocesses and analyzes the feature contribution of raw multi-source data to generate interpretable risk prediction reports. This solves the problem of decision-making processes that are difficult to interpret using machine learning models, and improves the efficiency and accuracy of risk prediction.

CN122241355APending Publication Date: 2026-06-19TONGDUN NETWORK TECH CO LTD
0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
TONGDUN NETWORK TECH CO LTD
Filing Date
2026-03-16
Publication Date
2026-06-19

AI Technical Summary

⚠Technical Problem

Existing machine learning models struggle to explain their decision-making processes in fraud detection, making it difficult for risk analysts to understand the basis of the model's judgments, and black-box models are unlikely to pass compliance reviews.

⚗Method used

An agent-based anti-fraud business processing system is adopted, including a Spark computing engine, an anti-fraud business database, and a business processing agent. By preprocessing the raw multi-source data, the contribution of features is determined and explanatory text is generated. Combined with intent parsing and reasoning fusion, an interpretable risk prediction report is constructed.

🎯Benefits of technology

It achieves interpretability in the fraud detection decision-making process, improves the accuracy of risk prediction reports, reduces computational burden, and meets compliance requirements.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

This disclosure relates to an agent-based anti-fraud business processing system, method, medium, and device, belonging to the field of big data processing technology. The system includes: a Spark computing engine, an anti-fraud business database, and a business processing agent. The Spark computing engine is used to calculate the original feature contribution of raw business data features to the model's predicted labels, and to generate original business explanation text based on the original feature contribution. The anti-fraud business database stores the raw business data features, raw user label data, raw feature contribution, and original business explanation text. The business processing agent performs reasoning and fusion on target business data, target user label data, target feature contribution, and target business explanation text to determine abnormal data features and causes of anomalies; and generates an interpretable risk prediction report based on the causes of anomalies and contextual features. This disclosure achieves interpretability of the decision-making process.
Need to check novelty before this filing date? Find Prior Art