Transaction Compliance Skip Rules for Real-Time Fraud Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing transaction compliance systems face overwhelming computational burdens due to the need to investigate all transactions in real-time, despite only a small fraction being suspicious, and parallel processing loses valuable sequential transaction information, while false positives are common.
Innovation Solution
A system utilizing machine learning to prioritize and execute only a subset of compliance tests relevant to real-time market conditions, skipping or delaying less relevant tests, and dynamically updating skip rules based on historical relevance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If all transactions are investigated in real-time to meet compliance regulations, then fraud detection reliability is improved, but computational burden increases excessively
Solution Approach 1:
The patent segments the compliance investigation process into two distinct pathways: real-time investigation for high-risk transactions and delayed/batched investigation for low-risk transactions. This segmentation allows the system to apply full computational resources only where necessary, thereby maintaining fraud detection reliability for critical cases while reducing overall computational burden through selective processing.
Solution Approach 2:
The system applies different levels of investigation intensity to different transactions based on their risk profile. High-risk transactions receive comprehensive real-time analysis with full computational resources, while low-risk transactions receive minimal or delayed analysis. This local differentiation of quality ensures reliable detection where needed while optimizing resource allocation across the entire transaction stream.
2Productivity
If parallel processing is used to reduce computational burden, then processing speed is improved, but sequential transaction information is lost
Solution Approach 1:
The patent implements a dynamic processing approach where the degree of parallelization adjusts based on transaction risk characteristics. For transactions requiring sequential analysis (where chronological information is critical), the system maintains serialized processing. For independent transactions, parallel processing is applied. This dynamic adaptation allows the system to optimize processing speed without universally sacrificing sequential information integrity.
Solution Approach 2:
The transaction stream is segmented into groups that can be processed in parallel while preserving necessary sequential relationships within each group. The system identifies independent transactions suitable for parallel processing and maintains chronological ordering for transactions where sequence matters, thereby achieving speed improvement without complete loss of sequential information.
3Measurement precision
If comprehensive compliance tests are executed on all transactions, then detection accuracy is improved, but false positive alerts increase
Solution Approach 1:
The system applies different levels of compliance testing to different transactions based on their risk profile. High-risk transactions undergo comprehensive testing with all compliance checks to ensure high detection accuracy. Low-risk transactions receive reduced or selective testing, which maintains adequate detection accuracy while significantly reducing false positive alerts generated by overly sensitive comprehensive checks on benign transactions.
Solution Approach 2:
Instead of applying excessive comprehensive testing to all transactions (which generates false positives), the system applies partial testing appropriately matched to transaction risk levels. This calibrated approach ensures sufficient detection accuracy for each transaction category while avoiding the harmful effect of false positives that result from unnecessary exhaustive testing of low-risk transactions.
Data Source
AI summary
A device, system and method for accelerated compliance testing for real-time detection of suspicious transactions in a transaction stream. A machine learning model may determine a global trend of cumulative transaction behavior based on all, a majority or a representative subset, of the stream of transactions. For each transaction in the stream, a skip rule engine may sort each security compliance test to be skipped or not skipped based on the test's relevance to the global trend of the cumulative transaction behavior. The not skipped security compliance tests may be executed in real-time for each transaction in the stream to generate real-time partial security assessments therefore comprising real-time suspicious transaction alerts. The remaining tests may be skipped in real-time and executed, at a time delay after transaction times, to generate time delayed supplemental security assessments therefore comprising time delayed suspicious transaction alerts to complete all compliance testing.