Post-Authorization Fraud Scoring Using Forward Transaction Velocities
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing fraud detection systems in payment card transactions fail to consider subsequent or near-term future transactions when evaluating a current transaction for fraud, leading to undetected fraudulent activities and increased network load.
Innovation Solution
A post-authorization modeling system that utilizes machine learning to re-score transactions by analyzing both past and future transactions using backward and forward velocities, enhancing the decision intelligence score to detect potential fraud.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional fraud detection systems evaluate transactions one at a time using only past data, then the system complexity remains low and processing speed is fast, but fraud detection accuracy deteriorates because future transactions cannot be considered
Solution Approach 1:
The system performs preliminary scoring of transactions using backward velocity (past data) before final authorization. This initial scoring allows the system to prepare fraud assessment results in advance, and then later update these scores with forward velocity data from subsequent transactions, thereby improving detection accuracy without requiring complete re-evaluation of all transactions
Solution Approach 2:
The system implements feedback by continuously monitoring subsequent transactions and using forward velocity to update the fraud scores of previously authorized transactions. When new transactions occur, the system feeds this information back to re-score earlier transactions, creating a closed-loop system that improves accuracy by incorporating future transaction patterns
2Measurement precision
If the system continuously monitors and re-scores all authorized transactions using future transactions, then fraud detection accuracy improves, but network load and computational resources deteriorate
Solution Approach 1:
Instead of re-scoring all authorized transactions continuously, the system applies partial action by focusing computational resources only on transactions that meet specific criteria (e.g., high-risk indicators, unusual patterns). This selective approach maintains improved fraud detection accuracy while significantly reducing the computational burden compared to universal continuous re-scoring
Solution Approach 2:
The system implements periodic monitoring and re-scoring of transactions rather than continuous evaluation. By checking transactions at intervals and triggering re-scoring based on specific events (new transactions, pattern changes), the system achieves improved fraud detection while consuming fewer computational resources compared to constant monitoring
Data Source
AI summary
A computing system for applying post-authorization modeling tools to a payment transaction for enhancing a decision intelligence (DI) score is disclosed. The computing system is configured to: (i) build a pre-authorization and a post-authorization model to analyze backward velocities and forward velocities; (ii) receive an authorization message for a current transaction; (iv) input into the pre-authorization model current transaction data and the backward velocities for the current transaction to output a DI score; (v) based upon the DI score, authorize the current transaction; (vi) continuously monitor a data feed for a predefined period of time after authorizing the current transaction; (viii) update a data register to include the forward velocities for any post-authorization transactions received during the predefined period of time; (ix) update the DI score for the current transaction by inputting the forward velocities into the post-authorization model to transmit the updated DI score to the merchant.