Dynamic Risk Weight Learning in Fraud Detection Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current fraud detection systems face challenges in accurately monitoring financial transactions and detecting fraudulent activities due to the complexity of large amounts of information and multiple parties involved, relying on deterministic methods that are less effective in dynamic scenarios.
Innovation Solution
A fraud detection system utilizing machine learning to determine and dynamically update risk weights in a relationship network, applying propagation models to compute risk by association with increased accuracy, incorporating data processing and risk models to evaluate party associations and update risk weights based on association scores.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If deterministic methods (static weights based on domain rules) are used to compute risk by association, then the system is simple to implement, but the accuracy of fraud detection deteriorates in dynamic scenarios
Solution Approach 1:
The patent transforms static risk weights into dynamic risk weights that are continuously updated based on real-time transaction data and learned patterns. The system adapts to changing fraud patterns by dynamically adjusting the importance of different relationship types and risk factors, allowing accurate detection in evolving scam scenarios while maintaining computational efficiency through incremental learning.
Solution Approach 2:
The system changes the parameters of risk assessment by introducing machine learning models that automatically adjust weight parameters based on data patterns. Instead of fixed domain rules, the system learns optimal parameter values from historical data, enabling accurate detection of complex fraud patterns while reducing manual configuration requirements.
2Measurement precision
If machine learning is applied to dynamically determine risk weights, then the accuracy of fraud detection is improved, but the complexity of the system increases
Solution Approach 1:
The system implements self-service through automated machine learning models that autonomously learn from data and adjust risk weights without manual intervention. The models automatically ingest transaction data, identify patterns, and update risk parameters, reducing the need for manual system configuration and expert intervention while maintaining high detection accuracy.
Solution Approach 2:
The system incorporates feedback loops where detection results and ground truth data are continuously fed back into the machine learning models to refine risk weight calculations. This feedback mechanism allows the system to learn from past performance and continuously improve accuracy while automating the tuning process, thereby managing complexity through self-optimization.
3Ease of repair
If static risk weights based on domain rules are used, then the system is easy to maintain, but the adaptability to new fraud patterns deteriorates
Solution Approach 1:
The system performs preliminary action by pre-training machine learning models on historical fraud data to establish baseline risk weights before deployment. This preliminary learning phase enables the system to quickly adapt to new fraud patterns when deployed, as the models are already equipped with general fraud detection capabilities that can be fine-tuned with minimal additional data, reducing ongoing maintenance requirements.
Data Source
AI summary
A computer-implemented system, method and computer program product for detecting fraud. The system and method receives data representing transacting parties where a party transacts with another party to define a relationship therebetween. The received data is incorporated into a relationship network graph comprising nodes that capture data about transacting parties and corresponding information associated with each transaction. A risk model is run that is configured to determine a risk weight for the relation between nodes associated with the transacting parties based on the data captured in the relationship network graph. Then there is determined a degree of a party's association score based on risk scores of nodes associated with the party and a defined relationship with another suspicious party. The system then dynamically updates the risk weight of the relation based on the party's association score.