Machine Learning Rule Generation for Real-Time Transaction Fraud Control

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Transaction systems face inefficiencies due to poorly performing transaction rules that either decline legitimate transactions or fail to detect fraudulent ones, leading to reduced performance and increased computing overhead.

Innovation Solution

Implement a machine learning-based system that detects trends in transaction data, generates optimized transaction rules, translates them for a live platform, and monitors their performance to ensure effective real-time decision-making.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If traditional transaction rules are used for real-time assessment, then the system can identify and decline potentially risky transactions, but overly restrictive rules result in declining legitimate transactions and reduced system effectiveness

Engineering Contradiction:

Improvetransaction assessment accuracyVSAvoidsystem effectiveness

Core Design Contradiction:

ReliabilityVSProductivity

Solution Approach 1:

The system dynamically adjusts transaction rule parameters based on statistical analysis of transaction data. By monitoring changes in transaction patterns and automatically updating rule parameters, the system adapts to evolving fraud patterns while maintaining appropriate thresholds to avoid false declines of legitimate transactions.

Inventive Principle:

Principle #35Parameter changes

Solution Approach 2:

The system implements continuous monitoring of transaction rule performance and uses this feedback to automatically refine and update rules. By analyzing the outcomes of transaction assessments and identifying patterns in both fraudulent and legitimate transactions, the system iteratively improves rule accuracy and reduces false positives.

Inventive Principle:

Principle #23Feedback

2Reliability

If transaction rules are defined based on offline analysis of historical data, then rules can be generated to improve loss prevention performance, but the conversion and deployment process creates bottlenecks that negatively affect system performance and efficiency

Engineering Contradiction:

Improveloss prevention performanceVSAvoidrule deployment time

Core Design Contradiction:

ReliabilityVSLoss of time

Solution Approach 1:

The system performs preliminary statistical analysis and rule generation continuously in the background using offline computational resources. By pre-processing transaction data and preparing optimized rules before they are needed for live assessment, the system eliminates deployment bottlenecks and enables rapid implementation of improved rules without affecting real-time performance.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system replaces manual rule conversion and deployment processes with automated machine learning models. These models directly translate offline analytical insights into executable transaction rules, eliminating the need for manual format conversion and reducing deployment time from days to minutes or seconds.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

3Reliability

If numerous transaction rules are applied to assess potential transactions in real-time, then the system can mitigate loss from fraudulent transactions, but the computing overhead and resources required increase significantly

Engineering Contradiction:

Improvefraud detection capabilityVSAvoidcomputing resources

Core Design Contradiction:

ReliabilityVSUse of energy by moving object

Solution Approach 1:

The system extracts and focuses computational resources on the most critical transaction assessment tasks by identifying and prioritizing high-risk transaction patterns. By using statistical analysis to pinpoint the most significant fraud indicators, the system applies complex rules selectively rather than uniformly to all transactions, reducing overall computing overhead while maintaining fraud detection effectiveness.

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The system applies a tiered approach to rule assessment where not all rules are evaluated for every transaction. Based on initial screening and transaction characteristics, the system selectively applies only the necessary subset of rules, avoiding the computational expense of evaluating every rule against every transaction while still maintaining comprehensive fraud detection coverage.

Inventive Principle:

Principle #16Partial or excessive action

Data Source

PatentUS20250245668A1Rule generation and management using machine learning

Publication Date: 2025.07.31 PAYPAL INC

AI summary

The disclosed computer-implemented method includes calculating, from transaction data, a statistical change in data entries corresponding to a type of transaction and modeling a transaction rule for normalizing the statistical change by changing an acceptance standard of the type of transaction. The method further includes activating the transaction rule to update a live database system for entering real-time data entries. Various other methods, systems, and computer-readable media are also disclosed.