Fraud Detection Rule Selection via Shapley Value Contribution
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing fraud detection systems in financial transactions face challenges in managing rules effectively, as local performance measures do not account for the overall contribution of rules to the system's performance, leading to potential missed frauds when rules with poor local performance but high overall contribution are removed.
Innovation Solution
A system that calculates the Shapley value of each rule to estimate its contribution to the overall performance of the rule set, allowing for the selection and management of a subset of rules that optimize the entire system's performance, including a rule evaluation module to generate evaluation reports and a rule selection module to automatically or manually select rules based on their Shapley values.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If rules are managed based on local performance measures (individual rule accuracy, backups, number of alerts), then the governance process becomes simpler and more straightforward, but rules with poor local performance but high overall contribution may be incorrectly suppressed, increasing fraud detection risk
Solution Approach 1:
The patent introduces an intermediary mechanism (Shapley value calculation module) that mediates between local rule performance and overall system performance. This intermediary computes the marginal contribution of each rule to the ensemble performance, allowing governance decisions to be based on accurate overall contribution rather than misleading local metrics, thus resolving the contradiction between governance simplicity and detection reliability
Solution Approach 2:
The patent replaces the mechanical system of direct local performance measurement with a computational system based on Shapley value theory. Instead of directly using local metrics (accuracy, backups, alerts) for governance, the system substitutes this with a game-theoretic approach that calculates each rule's fair share of the overall performance, eliminating the need for complex manual analysis while ensuring reliable fraud detection
2Reliability
If a large set of rules is used in the fraud detection system, then the overall performance and coverage of fraud detection is improved, but the complexity of rule set governance increases significantly
Solution Approach 1:
The patent implements a self-service mechanism where the system automatically evaluates and ranks rules based on their Shapley values without requiring manual intervention. The governance process becomes self-managing, with the system automatically identifying high-contribution rules and suppressing low-contribution ones, thus maintaining high fraud detection coverage while eliminating the complexity of manual rule set management
Solution Approach 2:
The patent changes the key parameter for rule evaluation from local performance metrics to Shapley values that represent overall contribution. This parameter change transforms the governance process, allowing a large number of rules to be managed efficiently by simply ranking them according to their computed Shapley values and applying threshold-based selection, thereby maintaining comprehensive fraud detection coverage without management complexity
Data Source
Figure 1
Figure 2a~2b
Figure 2c~3a
AI summary
The invention relates to a system and method for managing financial transactions, the system including at least one database containing the transaction information, at least one fraud detection device comprising at least one rule evaluation module and a rule selection module, connected to each other and comprising means for implementing the method, the evaluation module making it possible to calculate an estimate of the contribution of each rule of the set of rules relative to a parameter representing the overall performance of a set of rules stored in a first memory of the detection device, and an evaluation report file analysed by the selection module to select a subset of rules from among the evaluated rules of the set of rules, the selected rules being stored in a second memory of the fraud detection device to be used for transaction control.