Granted Patent B1

US 12,432,241 · App. 19/070,261 · Granted Sep 30, 2025

Inventors: Soumitri Naga Kolavennu (Plymouth, MN); Jayant Chawla (Minneapolis, MN); Varshini Sriram (San Jose, CA)

Assignee: U.S. Bank National Association

H04L63/1425H04L41/16H04L63/20

Abstract

Claims (188)

1. A method for improving transaction screening accuracy, comprising:

training agentic systems using historical false positive data;

dynamically adjusting screening thresholds and predefined criteria based on past outcomes; and

providing explanations for screening decisions to aid manual reviewers by:

generating rationales for recommendations based on specific screening criteria and threshold violations,

recording reason codes that map to identified risk patterns and compliance rules, and

storing the specific criteria and data elements that triggered each screening decision in a training database.

2. The method of claim 1 , further comprising:

integrating feedback loops to refine future screening recommendations and reduce false positives over time.

3. The method of claim 1 , further comprising:

analyzing feedback patterns across multiple time horizons;

tracking key metrics including false positive rates and processing times; and

initiating threshold recalibration procedures based on sustained performance patterns.

4. The method of claim 1 , further comprising:

maintaining performance histories for each agentic system by tracking detection accuracy and false positive rates;

selecting future agentic systems based on historical performance metrics and transaction risk profiles; and

adapting to emerging transaction behavior patterns while maintaining regulatory compliance through dynamic threshold adjustments.

5. The method of claim 1 , wherein training agentic systems comprises:

utilizing supervised learning techniques with labeled datasets;

incorporating reinforcement learning mechanisms; and

periodically retraining using updated datasets.

6. The method of claim 1 , further comprising:

implementing a temporal analysis framework;

evaluating threshold effectiveness across different time periods;

maintaining audit trails for regulatory compliance.

7. The method of claim 1 , further comprising:

implementing natural language processing models for entity name matching;

utilizing anomaly detection models for identifying suspicious patterns; and

adjusting model parameters based on advisory system feedback.

8. The method of claim 1 , further comprising:

analyzing both explicit feedback from advisory subsystems and implicit feedback from transaction patterns; and

automatically updating screening parameters based on real-time sanction list data feed updates and pattern detection algorithms.

9. The method of claim 1 , further comprising:

configuring a hierarchical rules engine to incorporate artificial intelligence models for transaction analysis;

executing cascading decision trees to evaluate specific criteria and decision-making steps;

generating evaluation outputs comprising:

detection accuracy metrics for each decision node;

identified patterns in false positive triggers; and

recommended threshold adjustments; and

automatically applying the recommended threshold adjustments to refine future screening decisions while maintaining audit trails for regulatory compliance.

10. The method of claim 1 , further comprising:

disaggregating entity traits in transacting parties and watchlist entities;

performing intelligent matching using multiple analytical tools.

11. The method of claim 1 , wherein dynamically adjusting screening thresholds comprises:

calculating optimal threshold levels using statistical analysis of historical transaction outcomes;

automatically tuning detection parameters based on false positive rates and processing efficiency metrics; and

implementing predictive models to adjust threshold values according to emerging risk patterns and transaction volumes.

12. A method for improving transaction screening accuracy, comprising:

training agentic systems using historical false positive data;

dynamically adjusting screening thresholds and predefined criteria based on past outcomes;

providing explanations for screening decisions to aid manual reviewers;

configuring a hierarchical rules engine to incorporate artificial intelligence models for transaction analysis;

executing cascading decision trees to evaluate specific criteria and decision-making steps;

generating evaluation outputs comprising:

detection accuracy metrics for each decision node;

identified patterns in false positive triggers; and

recommended threshold adjustments; and

automatically applying the recommended threshold adjustments to refine future screening decisions while maintaining audit trails for regulatory compliance.

13. The method of claim 12 , further comprising:

integrating feedback loops to refine future screening recommendations and reduce false positives over time.

14. The method of claim 12 , further comprising:

analyzing feedback patterns across multiple time horizons;

tracking key metrics including false positive rates and processing times; and

initiating threshold recalibration procedures based on sustained performance patterns.

15. The method of claim 12 , further comprising:

maintaining performance histories for each agentic system by tracking detection accuracy and false positive rates;

selecting future agentic systems based on historical performance metrics and transaction risk profiles; and

adapting to emerging transaction behavior patterns while maintaining regulatory compliance through dynamic threshold adjustments.

16. The method of claim 12 , wherein training agentic systems comprises:

utilizing supervised learning techniques with labeled datasets;

incorporating reinforcement learning mechanisms; and

periodically retraining using updated datasets.

17. The method of claim 12 , further comprising:

implementing a temporal analysis framework;

evaluating threshold effectiveness across different time periods;

maintaining audit trails for regulatory compliance.

18. The method of claim 12 , further comprising:

implementing natural language processing models for entity name matching;

utilizing anomaly detection models for identifying suspicious patterns; and

adjusting model parameters based on advisory system feedback.

19. The method of claim 12 , wherein providing the explanations comprises:

generating rationales for recommendations based on specific screening criteria and threshold violations;

recording reason codes that map to identified risk patterns and compliance rules; and

storing the specific criteria and data elements that triggered each screening decision in a training database.

20. The method of claim 12 , further comprising:

analyzing both explicit feedback from advisory subsystems and implicit feedback from transaction patterns; and

automatically updating screening parameters based on real-time sanction list data feed updates and pattern detection algorithms.

21. The method of claim 12 , further comprising:

disaggregating entity traits in transacting parties and watchlist entities;

performing intelligent matching using multiple analytical tools.

22. The method of claim 12 , wherein dynamically adjusting screening thresholds comprises:

calculating optimal threshold levels using statistical analysis of historical transaction outcomes;

automatically tuning detection parameters based on false positive rates and processing efficiency metrics; and

implementing predictive models to adjust threshold values according to emerging risk patterns and transaction volumes.

23. A method for improving transaction screening accuracy, comprising:

training agentic systems using historical false positive data;

dynamically adjusting screening thresholds and predefined criteria based on past outcomes;

providing explanations for screening decisions to aid manual reviewers;

disaggregating entity traits in transacting parties and watchlist entities; and

performing intelligent matching using multiple analytical tools.

24. The method of claim 23 , further comprising:

integrating feedback loops to refine future screening recommendations and reduce false positives over time.

25. The method of claim 23 , further comprising:

analyzing feedback patterns across multiple time horizons;

tracking key metrics including false positive rates and processing times; and

initiating threshold recalibration procedures based on sustained performance patterns.

26. The method of claim 23 , further comprising:

maintaining performance histories for each agentic system by tracking detection accuracy and false positive rates;

selecting future agentic systems based on historical performance metrics and transaction risk profiles; and

adapting to emerging transaction behavior patterns while maintaining regulatory compliance through dynamic threshold adjustments.

27. The method of claim 23 , wherein training agentic systems comprises:

utilizing supervised learning techniques with labeled datasets;

incorporating reinforcement learning mechanisms; and

periodically retraining using updated datasets.

28. The method of claim 23 , further comprising:

implementing a temporal analysis framework;

evaluating threshold effectiveness across different time periods;

maintaining audit trails for regulatory compliance.

29. The method of claim 23 , further comprising:

implementing natural language processing models for entity name matching;

utilizing anomaly detection models for identifying suspicious patterns; and

adjusting model parameters based on advisory system feedback.

30. The method of claim 23 , wherein providing the explanations comprises:

generating rationales for recommendations based on specific screening criteria and threshold violations;

recording reason codes that map to identified risk patterns and compliance rules; and

storing the specific criteria and data elements that triggered each screening decision in a training database.

31. The method of claim 23 , further comprising:

analyzing both explicit feedback from advisory subsystems and implicit feedback from transaction patterns; and

automatically updating screening parameters based on real-time sanction list data feed updates and pattern detection algorithms.

32. The method of claim 23 , further comprising:

configuring a hierarchical rules engine to incorporate artificial intelligence models for transaction analysis;

executing cascading decision trees to evaluate specific criteria and decision-making steps;

generating evaluation outputs comprising:

detection accuracy metrics for each decision node;

identified patterns in false positive triggers; and

recommended threshold adjustments; and

automatically applying the recommended threshold adjustments to refine future screening decisions while maintaining audit trails for regulatory compliance.

33. The method of claim 23 , wherein dynamically adjusting screening thresholds comprises:

calculating optimal threshold levels using statistical analysis of historical transaction outcomes;

automatically tuning detection parameters based on false positive rates and processing efficiency metrics; and

implementing predictive models to adjust threshold values according to emerging risk patterns and transaction volumes.

34. A method for improving transaction screening accuracy, comprising:

training agentic systems using historical false positive data;

dynamically adjusting screening thresholds and predefined criteria based on past outcomes by:

calculating optimal threshold levels using statistical analysis of historical transaction outcomes,

automatically tuning detection parameters based on false positive rates and processing efficiency metrics, and

implementing predictive models to adjust threshold values according to emerging risk patterns and transaction volumes; and

providing explanations for screening decisions to aid manual reviewers.

35. The method of claim 34 , further comprising:

integrating feedback loops to refine future screening recommendations and reduce false positives over time.

36. The method of claim 34 , further comprising:

analyzing feedback patterns across multiple time horizons;

tracking key metrics including false positive rates and processing times; and

initiating threshold recalibration procedures based on sustained performance patterns.

37. The method of claim 34 , further comprising:

maintaining performance histories for each agentic system by tracking detection accuracy and false positive rates;

selecting future agentic systems based on historical performance metrics and transaction risk profiles; and

adapting to emerging transaction behavior patterns while maintaining regulatory compliance through dynamic threshold adjustments.

38. The method of claim 34 , wherein training agentic systems comprises:

utilizing supervised learning techniques with labeled datasets;

incorporating reinforcement learning mechanisms; and

periodically retraining using updated datasets.

39. The method of claim 34 , further comprising:

implementing a temporal analysis framework;

evaluating threshold effectiveness across different time periods;

maintaining audit trails for regulatory compliance.

40. The method of claim 34 , further comprising:

implementing natural language processing models for entity name matching;

utilizing anomaly detection models for identifying suspicious patterns; and

adjusting model parameters based on advisory system feedback.

41. The method of claim 34 , wherein providing the explanations comprises:

generating rationales for recommendations based on specific screening criteria and threshold violations;

recording reason codes that map to identified risk patterns and compliance rules; and

storing the specific criteria and data elements that triggered each screening decision in a training database.

42. The method of claim 34 , further comprising:

analyzing both explicit feedback from advisory subsystems and implicit feedback from transaction patterns; and

automatically updating screening parameters based on real-time sanction list data feed updates and pattern detection algorithms.

43. The method of claim 34 , further comprising:

configuring a hierarchical rules engine to incorporate artificial intelligence models for transaction analysis;

executing cascading decision trees to evaluate specific criteria and decision-making steps;

generating evaluation outputs comprising:

detection accuracy metrics for each decision node;

identified patterns in false positive triggers; and

recommended threshold adjustments; and

automatically applying the recommended threshold adjustments to refine future screening decisions while maintaining audit trails for regulatory compliance.

44. The method of claim 34 , further comprising:

disaggregating entity traits in transacting parties and watchlist entities;

performing intelligent matching using multiple analytical tools.

Assignments (1)

Continuity (1)

References Cited (10)