Multi-stage filtering for fraud detection with account event data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional fraud detection systems are inefficient, leading to both missed fraudulent transactions and excessive false-positive identifications, causing financial losses and consumer frustration, as they analyze all transactions equally without distinguishing between high-risk and low-risk activities.
Innovation Solution
A multi-stage filtering system that analyzes financial transactions using both transaction data and account event data, such as password changes and financial product orders, to identify high-risk transactions, thereby reducing unnecessary scrutiny and resource allocation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional fraud detection systems analyze all transactions equally, then they can detect fraudulent transactions, but they produce excessive false-positive identifications and increase data storage costs
Solution Approach 1:
The patent segments the fraud detection process into multiple stages: initial filtering using account event data, intermediate filtering using transaction data, and final fraud analysis. This segmentation allows the system to handle different types of transactions with appropriate scrutiny levels, reducing false positives while maintaining detection accuracy.
Solution Approach 2:
The patent applies different filtering criteria and analysis depths to different transactions based on their risk characteristics. High-risk transactions receive comprehensive multi-stage analysis, while low-risk transactions are quickly filtered or processed with minimal scrutiny, optimizing resource allocation and reducing false positives.
2Reliability
If conventional fraud detection systems apply strict filtering to eliminate fraudulent transactions, then they can prevent fraud, but they decline legitimate transactions causing loss of revenue
Solution Approach 1:
The patent performs preliminary filtering using account event data before conducting detailed fraud analysis on transaction data. This preliminary action identifies high-risk accounts and transactions in advance, allowing the system to apply stricter scrutiny only where necessary and process legitimate transactions efficiently.
Solution Approach 2:
The patent applies partial filtering by focusing computational resources on specific high-risk transactions identified through account event data, rather than applying uniform strict filtering to all transactions. This selective approach maintains fraud prevention capability while minimizing impact on legitimate transactions.
3Measurement precision
If fraud detection systems analyze every transaction in detail, then they can identify fraudulent patterns, but they consume excessive computational resources and time
Solution Approach 1:
The patent divides fraud detection into sequential stages with increasing analysis depth. Stage 1 uses account event data for quick risk assessment, Stage 2 applies transaction data filtering, and Stage 3 performs detailed fraud pattern analysis only on suspicious transactions, optimizing the balance between detection precision and processing efficiency.
Solution Approach 2:
The patent applies detailed fraud pattern analysis only to transactions that fail earlier filtering stages, rather than analyzing every transaction equally. This partial application of intensive analysis maintains detection precision for high-risk cases while significantly improving overall processing efficiency.
Data Source
AI summary
A multi-stage filtering process and system for fraud detection is disclosed. The process includes one or more preliminary filtration stages followed by one or more additional filtration stages that provide for enhanced screening for fraudulent activity. In one such embodiment the preliminary filtration provides for evaluating financial transactions based on financial transaction attribute data and secondary filtration provides for further evaluating the financial transactions based on a customer's account event data. Over a plurality of transactions, a portion of the transactions are cleared for processing (e.g., deemed not likely fraudulent or of too low value to continue processing) after each filtration stage. As such, acceptable transactions are not unnecessarily scrutinized. In specific embodiments, filtration is based on financial transaction data and account event data.