Fraud Alert Tuning with Customer Feedback and Machine Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing fraud detection systems generate a high number of false positive alerts, leading to significant resource drain and customer annoyance, despite their intention to capture fraudulent transactions.
Innovation Solution
Implement a computer-implemented method using customer feedback and machine learning to modify rules-based engines, determining the reasons for false positives and adjusting detection criteria accordingly to reduce future false alerts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a low threshold is used to trigger fraud alerts to ensure most fraud instances are captured, then fraud detection coverage is improved, but the number of false positive alerts increases
Solution Approach 1:
The system implements feedback loops where customer responses to fraud alerts (confirming fraud or indicating false positive) are fed back into the machine learning model. This allows the system to learn from actual outcomes and continuously refine its detection threshold and criteria, resolving the contradiction by adapting to real-world performance data.
Solution Approach 2:
The machine learning model dynamically adjusts detection parameters and thresholds based on learned patterns from training data. Instead of using fixed low thresholds that generate false positives, the system optimizes parameters to balance detection coverage with false positive reduction, achieving both goals simultaneously.
2Reliability
If numerous fraud alerts are generated to capture fraud instances, then fraud detection sensitivity is improved, but human resource drain increases
Solution Approach 1:
The system enables self-service fraud detection by allowing customers to directly confirm or deny alerts through mobile devices. This automated customer verification process eliminates the need for manual investigation of each alert, reducing human resource consumption while maintaining high detection sensitivity through the initial broad alert generation.
Solution Approach 2:
Customer feedback on alert accuracy is systematically collected and used to retrain the machine learning model. This creates a self-improving system that becomes increasingly accurate over time, allowing the system to maintain high sensitivity while progressively reducing the volume of alerts requiring human review.
3Reliability
If numerous fraud alerts are generated to capture fraud instances, then fraud detection sensitivity is improved, but customer annoyance increases
Solution Approach 1:
Customers take control of the fraud verification process through mobile device interactions, confirming or denying alerts at their convenience. This self-service approach reduces the perceived burden and annoyance by empowering customers rather than having alerts imposed upon them without agency.
Solution Approach 2:
The system optimizes alert parameters and delivery methods based on customer behavior patterns learned through machine learning. By adjusting when and how alerts are delivered, the system maintains high detection sensitivity while minimizing disruption and annoyance to legitimate customers.
Data Source
AI summary
A method of reducing a future amount of electronic fraud alerts includes receiving data detailing a financial transaction, inputting the data into a rules-based engine that generates an electronic fraud alert, transmitting the alert to a mobile device of a customer, and receiving from the mobile device customer feedback indicating that the alert was a false positive or otherwise erroneous. The method also includes inputting the data detailing the financial transaction into a machine learning program trained to (i) determine a reason why the false positive was generated, and (ii) then modify the rules-based engine to account for the reason why the false positive was generated, and to no longer generate electronic fraud alerts based upon (a) fact patterns similar to fact patterns of the financial transaction, or (b) data similar to the data detailing the financial transaction, to facilitate reducing an amount of future false positive fraud alerts.