Property Graph Data Synthesis for Fraud Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for generating pseudo-realistic data for training machine learning classifiers to detect fraudulent events are inefficient and require complex correlation tables, leading to imbalanced datasets that result in models insensitive to fraudulent activities due to the rarity of such events in real-world data.
Innovation Solution
A system and method that transforms input data into a property graph, allowing for the generation of case-based data by modifying network elements and properties while preserving calculated statistics, using techniques such as minority over-sampling and machine learning clustering to create synthetic datasets that mimic real-world scenarios, thereby balancing the dataset and improving model sensitivity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If real world data is used for training, then data authenticity is improved, but dataset balance deteriorates due to rarity of fraudulent events
Solution Approach 1:
The patent creates synthetic copies of fraudulent events by transforming real fraudulent cases into property graphs and generating modified versions through network element changes. This copying approach preserves the authentic characteristics of fraudulent patterns while generating sufficient training samples to balance the dataset.
Solution Approach 2:
The system modifies parameters of network elements in property graphs (such as entity attributes, link properties, and network structure) to generate varied synthetic fraudulent cases. By changing parameters like transaction amounts, timestamps, and entity relationships, the system creates diverse yet realistic fraudulent scenarios.
2Ease of manufacture
If Bayesian networks are used for data generation, then data synthesis capability is improved, but system complexity increases due to correlation tables
Solution Approach 1:
The patent replaces the complex mechanical system of Bayesian networks and correlation tables with a graph-based transformation approach. By using property graphs and network element modifications, the system achieves data synthesis without requiring gigantic correlation tables, significantly reducing system complexity.
Solution Approach 2:
The system segments the data generation process into discrete steps: transforming input data to property graphs, identifying network elements for modification, making controlled changes to create variations, and generating output datasets. This segmentation simplifies the overall process compared to monolithic Bayesian network approaches.
3Quantity of substance
If SMOTE technique is used for over-sampling, then minority class representation is improved, but data quality deteriorates due to synthetic nature
Solution Approach 1:
Instead of simple interpolation like SMOTE, the patent creates synthetic samples by copying and transforming real fraudulent case structures into property graphs, then modifying network elements to generate variations that preserve authentic fraudulent patterns while maintaining data quality.
Data Source
AI summary
System and method for generating case-based data including receiving, input data describing an event of interest; if the input data is not in a format of a property graph then transforming the input data into a first property graph describing the event of interest and representing a first network, wherein the first property graph includes a plurality of network elements and properties of at least some of the plurality of network elements, wherein the network elements include entities and links describing relationships between the entities; changing a network element in the first property graph to create a second property graph of a new network; and using the second property graph as the case-based data. New properties may be generated for the entities and links.