Distributed Event Representations for Fraud Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current systems lack efficient methods for accurately and dynamically identifying fraud risk and behavior anomalies in large datasets of user interactions, particularly in real-time, due to the complexity and volume of event data from various sources.
Innovation Solution
The implementation of artificial intelligence-guided monitoring systems that utilize machine learning and distributed representations of event data to generate vectors for similarity analysis, allowing for the identification of behavior anomalies by comparing candidate events to historical data, including the use of neural networks and language modeling to reduce data complexity and enhance similarity calculations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional methods are used to analyze user event data for fraud detection, then the system can process event data, but the accuracy and efficiency of identifying fraud risk and behavior anomalies deteriorates due to data complexity and volume
Solution Approach 1:
The patent transforms event data from its original complex format into distributed vector representations, changing the parameter space from discrete event attributes to continuous vector embeddings. This transformation enables more effective similarity calculations and anomaly detection by mapping events to a lower-dimensional space where patterns are more discernible.
Solution Approach 2:
The patent introduces distributed representations as an intermediary layer between raw event data and fraud detection analysis. These vector representations serve as a mediator that captures the essential characteristics of events while reducing complexity, enabling more accurate similarity comparisons without directly processing the full complexity of original event data.
2Measurement precision
If distributed representations are used to reduce data complexity, then fraud detection accuracy improves, but data storage requirements increase due to vector representations
Solution Approach 1:
The patent transitions event data from a high-dimensional discrete attribute space to a lower-dimensional continuous vector space. By projecting events into this compressed dimensional space, the system reduces storage requirements while preserving the essential patterns needed for accurate fraud detection through similarity measurements.
3Loss of time
If real-time analysis of large datasets is performed, then fraud detection timeliness improves, but processing speed deteriorates due to computational complexity
Solution Approach 1:
The patent pre-computes distributed vector representations for historical event data and stores them in advance. When a new event needs to be analyzed for fraud, the system only needs to compute its vector representation and compare it against the pre-processed historical vectors, significantly reducing the computational time required for real-time detection while maintaining high processing speed.
Data Source
AI summary
The features relate to artificial intelligence directed detection of user behavior based on complex analysis of user event data including language modeling to generate distributed representations of user behavior. Further features are described for reducing the amount of data needed to represent relationships between events such as transaction events received from card readers or point of sale systems. Machine learning features for dynamically determining an optimal set of attributes to use as the language model as well as for comparing current event data to historical event data are also included.