Fraud Detection Ruleset Evaluation via Partitioned Historical Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for evaluating and deploying new fraud detection rulesets in electronic transactions are slow, leading to a 90-day delay between finalization and deployment, making them outdated due to rapidly evolving fraud tactics.
Innovation Solution
A computer-implemented method that allows for the evaluation of fraud detection rulesets against historical transaction data within a substantially shorter timeframe, using concurrent execution and partitioned data structures to improve processing speed and accuracy, enabling quicker deployment of improved fraud detection strategies.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If new fraud detection rulesets are trialed concurrently with the current champion ruleset using real-world transaction data, then the accuracy and reliability of fraud detection can be improved, but the deployment time increases significantly causing a 90-day delay
Solution Approach 1:
The patent applies preliminary action by evaluating new rulesets against historical transaction data before deploying them to live traffic. This allows the system to pre-assess the performance of candidate rulesets using past data, so that when deployment is decided, the ruleset is already validated and ready for immediate implementation, eliminating the 90-day delay while maintaining accuracy improvements
Solution Approach 2:
The patent segments the evaluation process into distinct phases: historical data evaluation using partitioned data structures, and live traffic trialing. By dividing the transaction data into partitions that can be independently evaluated, the system can efficiently assess multiple rulesets in parallel against historical data, then proceed to controlled live deployment only for the most promising candidates, significantly reducing overall deployment time
2Adaptability or versatility
If fraud detection rulesets are updated frequently to keep pace with evolving fraud tactics, then the detection effectiveness improves, but the complexity of managing and testing multiple rulesets increases
Solution Approach 1:
The patent implements dynamics by creating a flexible, adaptive framework for ruleset management where candidate rulesets can be dynamically added, evaluated, and deployed based on their performance. The system continuously monitors fraud patterns and can rapidly iterate through multiple ruleset versions, promoting effective ones to champion status while demoting or removing ineffective ones, allowing the system to adapt quickly to evolving fraud tactics without manual intervention
Solution Approach 2:
The patent uses copying by maintaining multiple candidate rulesets that can be evaluated in parallel against historical data. Instead of manually creating and testing each ruleset variant, the system can generate candidate rulesets by copying and modifying existing champion rulesets, then automatically evaluate them using the partitioned historical data structure, significantly reducing the complexity of managing multiple ruleset versions
3Measurement precision
If the evaluation process uses a large volume of historical transaction data to ensure accurate assessment, then the measurement precision of ruleset performance improves, but the processing time increases
Solution Approach 1:
The patent applies segmentation by dividing the large historical transaction data into multiple partitions organized in a data structure that enables efficient access. Each partition can be independently processed by different evaluation threads or processes, allowing parallel evaluation of multiple rulesets against the same historical data without interference. This segmentation maintains measurement precision by ensuring each ruleset is evaluated against the complete data set while dramatically improving evaluation speed through parallel processing
Solution Approach 2:
The patent introduces another dimension by organizing historical transaction data in a partitioned data structure that adds an organizational dimension to the data. This structure allows the evaluation process to access and process data from multiple dimensions simultaneously - temporal (historical periods), categorical (transaction types), and spatial (partition locations) - enabling efficient parallel processing while maintaining comprehensive evaluation coverage
Data Source
AI summary
This present disclosure relates to computer-implemented methods for use in electronic fraud detection systems, and to data processing apparatus, computer programs, and computer readable storage media for performing the same. One such computer-implemented method comprises accessing a data structure comprising a plurality of transaction records; and executing one or more decision processes sharing a common ruleset, each decision process being configured to use the ruleset to output fraud decisions for transaction records, wherein a fraud decision for a transaction record is a prediction based on the ruleset as to whether the transaction to which the transaction record relates is fraudulent; wherein for each decision process, executing the decision process comprises: reading one or more transaction records for a predetermined historical period from the data structure; and determining, using the ruleset, fraud decisions for the one or more transaction records read from the data structure.