Fraud Detection System Using Transaction Action Clustering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional fraudulent financial transaction detection systems are insufficiently accurate as they do not effectively differentiate between transaction details corresponding to various transaction action types, leading to inadequate detection of potentially fraudulent requests.
Innovation Solution
A method and system that collect transaction details from financial institutions, cluster users into groups based on transaction actions like Buy, Sell, Send, Receive, Deposit, Withdrawal, External Send, and External Receive, and classify users into risk groups using a machine learning module trained to analyze transaction patterns, determining risk scores for each group to identify potentially fraudulent activities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional fraud detection systems use general transaction request details and use patterns, then the system is simple to operate, but the detection accuracy is insufficient
Solution Approach 1:
The patent segments users into multiple clusters based on transaction action types (e.g., Buy, Sell, Send, Receive, Deposit, Withdrawal). Each cluster is analyzed separately with tailored risk assessment, enabling more accurate detection without requiring a completely complex new system architecture.
Solution Approach 2:
The patent introduces a new dimension of classification by categorizing users according to their transaction action types. This additional dimensional approach to analyzing transaction patterns enhances detection accuracy without fundamentally complicating the underlying detection mechanism.
2Measurement precision
If the system analyzes all transaction details without differentiation, then the system is easy to implement, but false positives increase
Solution Approach 1:
By segmenting users into distinct clusters based on transaction action types, the system reduces false positives through targeted analysis. Each cluster's specific patterns are evaluated separately, avoiding the false positives that arise from applying uniform thresholds to diverse transaction types.
Solution Approach 2:
The patent applies local quality by setting different risk thresholds and analysis parameters for different user clusters. Each cluster receives customized detection parameters appropriate to its transaction patterns, improving precision without requiring complete system redesign.
3Adaptability or versatility
If the system uses rule-based threshold detection, then the system is simple to operate, but it cannot effectively differentiate between various transaction action types
Solution Approach 1:
The system segments transaction detection into multiple clusters based on action types. Each cluster can have its own risk assessment rules and thresholds, enabling effective differentiation between transaction types while maintaining operational simplicity through automated clustering.
Solution Approach 2:
The patent creates a multi-functional detection system that handles multiple transaction action types (Buy, Sell, Send, Receive, Deposit, Withdrawal) within a unified framework. The same basic detection infrastructure serves multiple purposes by adapting to different user clusters.
Data Source
AI summary
Disclosed is a method and system for detecting a fraudulent financial transaction including collecting, by processing circuitry, transaction details from a financial institution, classifying, by the processing circuitry, each of a plurality of users into a respective set of groups among a plurality of groups for each of a plurality of transaction action types, the plurality of users corresponding to the transaction details, and determining, by the processing circuitry, whether a first user among the plurality of users is in a risk group based on a first set of groups among the plurality of groups into which the first user is classified.