Alert Classifier With Voting Engine For Fraud Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current automated monitoring systems for financial transactions face inefficiencies in linking and managing large volumes of alerts, particularly in identifying and investigating fraudulent activities, due to the lack of effective automated systems for exact matching and linking of alerts across multiple detection engines.
Innovation Solution
A monitoring system that includes an interface for receiving source alerts, a database of historical events, and a classifier with multiple match methods and a voting engine to classify alerts by linking them to historical events, using weighted outputs to determine links and improve accuracy over time through feedback mechanisms.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual analysis and linking of alerts is used, then accuracy of alert classification can be maintained, but productivity and time efficiency deteriorate significantly
Solution Approach 1:
The system enables automated self-linking of alerts through the classifier component that automatically associates alerts with historical events and other alerts based on matching criteria, eliminating the need for manual investigation and linking operations
Solution Approach 2:
The patent replaces manual mechanical processes of alert investigation and linking with an automated computational system comprising a classifier, match methods, and voting engine that perform linking operations algorithmically
2Productivity
If automated linking systems are implemented, then productivity improves, but device complexity increases
Solution Approach 1:
The system divides the complex alert linking function into separate components: a classifier for classification, multiple match methods for different matching criteria, a voting engine for aggregation, and a linking function for final link creation, making the overall system more manageable
Solution Approach 2:
The classifier component serves multiple functions including classification, linking to historical events, linking to current alerts, and generating output alerts, reducing the need for separate dedicated components for each function
3Measurement precision
If multiple match methods are used in the classifier, then measurement precision of alert matching improves, but device complexity increases
Solution Approach 1:
Multiple match methods are combined within a single classifier component, where each match method processes different aspects of alert data and their results are aggregated through the voting engine to produce a comprehensive matching outcome
Solution Approach 2:
The voting engine acts as an intermediary between multiple match methods and the final linking decision, aggregating and weighting the results from different match methods to produce a unified output that balances the complexity of multiple methods with a single decision mechanism
Data Source
AI summary
A monitoring system (1) comprises an interface (2) for receiving source alerts from at least one detection engine, a database (7) of historical events; and a classifier (3) for classifying received source alerts by linking a source alert with an historical event or a current source alert to provide a link, and providing said link as an output alert. The classifier comprises match methods (9) for processing source alerts and generating a score for extent of matching of a source alert with an historical event or current source alert, a voting engine (4) for weighting scores from the match methods (9), and a linking function (6) for determining that there is a link if a combination of the weighted outputs of a plurality of match methods exceeds a threshold. At least some match methods (9) are each associated with a specific field of a source alert such as a numerical value field or a name field of a source alert. A feedback function (6) notifies a case management system (5) of links, and the voting engine (4) receives from the case management system (5) feedback (11) of success of each match method (9), and adjusts match method weights (12) accordingly.