Real-Time Fraud Machine Learning Module Distributed Parallel Execution
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional fraud detection systems are inefficient in real-time transaction processing due to reliance on centralized mainframe computers handling single data sources, leading to delayed updates and increased fraud losses for card issuers.
Innovation Solution
Implementing a real-time fraud machine learning module (RTFMLM) with a distributed parallel architecture using proprietary machine learning algorithms, open-source software, and frameworks like PMML, Java Spring Boot, Cassandra, and Kafka, allowing for simultaneous execution of multiple models and frequent updates without manual recoding.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional fraud models are deployed on centralized mainframe computers, then system stability is maintained, but real-time fraud detection capability deteriorates
Solution Approach 1:
The patent segments the centralized mainframe system into a distributed architecture where multiple computing nodes process fraud detection independently. Each node can handle specific data sources or model types, enabling parallel processing that maintains system stability while achieving real-time detection capabilities through distributed computation.
Solution Approach 2:
The patent transitions from a single-dimensional centralized processing model to a multi-dimensional distributed architecture. This includes adding spatial distribution (multiple nodes), temporal dimension (real-time streaming processing), and functional dimension (multiple model types running simultaneously), thereby resolving the contradiction between stability and speed.
2Device complexity
If conventional fraud models handle single data sources, then processing simplicity is maintained, but detection accuracy deteriorates
Solution Approach 1:
The patent creates a universal fraud detection platform that can handle multiple data sources (transaction data, customer profiles, merchant information, device data) through a unified architecture. The system uses standardized data ingestion pipelines and configurable model frameworks that can process diverse data types without requiring separate processing systems for each data source.
Solution Approach 2:
The patent merges multiple data sources and model types into a unified fraud detection system. By combining transactional data, historical data, and real-time streaming data through integrated processing pipelines, the system achieves comprehensive detection accuracy while maintaining architectural simplicity through unified management.
3Measurement precision
If fraud models are manually recoded for updates, then model precision can be improved, but update frequency deteriorates
Solution Approach 1:
The patent implements self-service capabilities through automated model training pipelines that can retrain models using new data without manual recoding. The system includes automated feature engineering, model validation, and deployment workflows that enable frequent updates while maintaining or improving model precision through continuous learning from new transaction patterns.
Solution Approach 2:
The patent creates a dynamic model update system where models can be frequently retrained and deployed based on incoming data. The architecture supports version control, A/B testing, and gradual rollout of new models, enabling rapid adaptation to emerging fraud patterns while maintaining precision through systematic validation processes.
4Power
If expensive mainframe computers are used, then processing power is sufficient, but operational cost deteriorates
Solution Approach 1:
The patent replaces expensive mainframe computers with commodity hardware that can be deployed in distributed configurations. The system uses standard servers, cloud instances, or even containerized environments that are less expensive individually but provide sufficient aggregate processing power through parallelization and scalability.
Solution Approach 2:
The patent changes the operational parameters from centralized high-power processing to distributed moderate-power processing. By utilizing multiple less powerful nodes working in parallel, the system achieves equivalent or superior processing capacity while reducing operational costs through commodity hardware and efficient resource utilization across the distributed network.
Data Source
AI summary
Various methods, apparatuses, and media for implementing a fraud machine learning model execution module are provided. A processor generates a plurality of machine learning models. The processor generates historical aggregate data based on prior transaction activities of a customer from a plurality of databases for transactions. The processor also tracks activities of the customer during a new transaction authorization process and generates a transaction data; integrates the transaction data with the historical aggregate data; executes each of said machine learning models using the integrated transaction data and the historical aggregate data to generate a fraud score and stores the fraud score into the memory; and determines whether the new transaction is fraudulent based on the generated fraud score.