Merchant Correspondence Matrix for Out-of-Pattern Transaction Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing fraud detection systems for payment card transactions often fail to identify out-of-pattern behavior due to limited access to spending pattern data and merchant network data, leading to inconsistent fraud detection outcomes and delayed reporting of fraudulent transactions.
Innovation Solution
An inverse recommender system that generates a merchant correspondence matrix based on historical transaction data to determine an inverse recommender score for new transactions, identifying out-of-pattern behavior by analyzing merchant associations inferred from a cardholder's transaction history.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If aggregated spending pattern data at industry or merchant level is used for fraud detection, then fraud detection coverage is improved, but data privacy and security are worsened
Solution Approach 1:
The patent segments fraud detection into two layers: (1) aggregated industry/merchant level patterns for general fraud detection, and (2) individual cardholder transaction history for personalized out-of-pattern detection. This segmentation allows both broad coverage and privacy protection by keeping individual data localized while still benefiting from aggregate insights.
Solution Approach 2:
The patent applies local quality by using individualized cardholder transaction history data specifically for detecting out-of-pattern behavior, while using aggregated data for general fraud patterns. Each data type is applied where it provides the most value, with individual data protecting privacy while enabling personalized fraud detection.
2Loss of information
If limited access to spending pattern data and merchant network data is used, then data security is improved, but fraud detection accuracy is worsened
Solution Approach 1:
The patent performs preliminary action by building and maintaining a merchant correspondence matrix in advance that captures relationships between merchants based on cardholder transaction patterns. This pre-computed structure enables accurate fraud detection without needing real-time access to extensive raw data, improving both security and accuracy.
Solution Approach 2:
The patent introduces a merchant correspondence matrix as an intermediary data structure that mediates between limited access to raw spending data and the need for accurate fraud detection. This matrix captures essential patterns without requiring direct access to detailed individual transaction data, thus maintaining security while enabling accurate detection.
3Productivity
If traditional fraud scoring processes are used, then processing speed is improved, but detection of out-of-pattern behavior is worsened
Solution Approach 1:
The patent performs preliminary action by pre-computing the merchant correspondence matrix that encodes merchant relationships and cardholder patterns. During transaction processing, this pre-computed structure enables rapid out-of-pattern detection without requiring complex real-time analysis, thus maintaining both speed and accuracy.
Data Source
AI summary
An inverse recommender system for detecting out-of-pattern payment transactions includes a memory device and a processor programmed to receive transaction data. The transaction data corresponds to historical payment transactions between account holders and merchants. The processor is programmed to generate a merchant correspondence matrix including the merchants and counters indicating the number of historical payment transactions between merchant pairs of the merchants and the account holders. The processor is programmed to store the merchant correspondence matrix in a memory device linking the merchant pairs to each account holder. The processor receives additional transaction data associated with a new payment transaction between an account holder and a merchant, and to generate an inverse recommender score for the new payment transaction based on the account holder's historical payment transaction data. The account holder's historical payment transaction data includes historical payment transaction data associated with the merchants visited by the account holder.