Jointly-Trained ML Models for Transaction Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Machine learning-based transaction processing systems face inefficiencies due to sub-optimal control actions, leading to resource scarcity, fraudulent transactions, and increased computing and networking overhead, necessitating more efficient configuration methods.
Innovation Solution
The implementation of jointly-trained integrated machine learning models that optimize control actions by maximizing an objective function based on the probability of matching target actions, including risk prediction, routing, retry, and re-presentment decisions, to improve transaction processing efficiency and reduce resource requirements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If multiple independently implemented machine learning models are used at each stage of transaction processing, then comprehensive control actions can be taken, but computing resources and processing time increase significantly
Solution Approach 1:
The patent combines multiple independently implemented machine learning models into a single integrated machine learning system that performs all control actions (risk assessment, routing, retry, re-presentment) through unified joint training, reducing system complexity while maintaining comprehensive control capabilities
Solution Approach 2:
The integrated machine learning system performs multiple functions (risk model, routing model, retry model, re-presentment model) within a single unified framework, allowing one system to handle all control actions across different transaction processing stages
2Reliability
If overly restrictive control actions are taken on transactions, then fraudulent transactions are blocked, but legitimate transactions are rejected and system efficiency decreases
Solution Approach 1:
The patent uses joint training to dynamically optimize decision parameters across multiple control actions, adjusting the strictness of control actions based on learned patterns from transaction data, thereby balancing fraud detection with maintaining transaction throughput
Solution Approach 2:
The integrated system uses feedback from joint training on historical transaction outcomes to continuously improve control action decisions, learning from both fraudulent and legitimate transaction patterns to optimize the balance between blocking fraud and allowing legitimate transactions
3Reliability
If sub-optimal control actions are taken resulting in rejection of legitimate transactions or allowing fraudulent ones, then further processing is required, but this imposes overhead on the system and consumes valuable computing resources
Solution Approach 1:
The patent performs preliminary joint training of all control action models together before actual transaction processing, pre-optimizing decision parameters so that during live operation, control actions can be executed efficiently without requiring extensive additional processing or corrective actions
Data Source
AI summary
A device in a data processing system for training machine learning models receives a transaction and forwards it to at least one of a plurality of integrated control action models that use outputs of one model as inputs to other models. The models are machine learning models jointly trained for taking each control action of a plurality of control actions on the transaction to maximize an objective function based on probabilities of the control actions matching corresponding target control actions. The machine learning models include a risk model that outputs risk prediction information for a first control action that indicates whether or not to initiate processing of the transaction. The device further receives the risk prediction information from the risk model, and executes at least the first control action based on the risk prediction information.