Dynamic Variable Fee Assignment for Cross-Border Transactions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current solutions for calculating variable fees for cross-border transactions fail to account for transaction-specific variables such as category, region, and amount, leading to suboptimal fee calculations and inadequate visibility for issuers.
Innovation Solution
Implementing scenario-specific machine learning models that rank transactions based on a ratio of maximum approved amounts to fraud likelihood, assigning them to tiers, and using corresponding models to dynamically compute and assign variable fees in real-time, optimizing profit for issuers while improving customer experience.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If scenario-specific machine learning models are implemented to dynamically compute variable fees, then fee calculation accuracy and issuer profit are improved, but device complexity and computational resources increase
Solution Approach 1:
The patent segments transactions into different scenarios (e.g., cross-border, domestic, high-risk, low-risk) and applies different machine learning models to each scenario. This segmentation allows the system to use complex models only where necessary while simpler models handle routine transactions, thereby improving fee calculation accuracy for specific scenarios without requiring all transactions to undergo complex processing.
Solution Approach 2:
The system dynamically changes parameters such as fee percentages, approval thresholds, and risk weights based on transaction characteristics and scenario classification. By adjusting these parameters according to the specific scenario and real-time conditions, the system achieves high calculation accuracy without needing a single overly complex model for all cases.
2Productivity
If real-time dynamic fee computation is performed for each transaction, then fee optimization and profit maximization are improved, but processing time and computational power increase
Solution Approach 1:
The system performs preliminary actions by pre-training machine learning models on historical transaction data and pre-establishing scenario classifications and fee structures. This allows the models to be deployed and execute quickly during real-time transaction processing, as the heavy computational work of model training and scenario definition has already been completed in advance.
Solution Approach 2:
The patent uses historical transaction data and patterns as copies to predict optimal fees for current transactions. By analyzing and replicating successful fee structures from historical data, the system can quickly determine appropriate fees without performing exhaustive real-time analysis, thus maintaining speed while optimizing profit.
3Adaptability or versatility
If multiple scenario-specific models are deployed, then adaptability to different transaction types is improved, but system complexity and model management difficulty increase
Solution Approach 1:
The patent implements a universal framework that can accommodate multiple scenario-specific models through a common interface and standardized data structures. The core system remains universal in handling different transaction types, while allowing specialized models to be plugged in for specific scenarios. This multi-functionality approach enables the system to adapt to various transaction types without requiring separate management systems for each model.
Solution Approach 2:
The system introduces an intermediary layer (the scenario classification and model routing mechanism) that sits between transaction input and the various ML models. This intermediary automatically routes transactions to the appropriate scenario-specific model based on transaction characteristics, thereby managing the complexity of multiple models through a centralized coordination layer rather than requiring direct management of each individual model.
4Measurement precision
If feedback loops are implemented for continuous model training, then model accuracy and performance are improved, but computational resources and processing overhead increase
Solution Approach 1:
The system implements periodic action by training models on feedback at scheduled intervals (e.g., daily, weekly, or monthly) rather than continuously for every single transaction. This periodic retraining maintains model accuracy by incorporating new data and patterns while avoiding the excessive computational energy that would result from continuous real-time retraining on every transaction.
Data Source
AI summary
Systems and methods dynamically compute and assign a variable fee for cross-border transactions. Raw data associated with a pending transaction is prepared for analysis by a scenario-specific model. The pending transaction is assigned to a scenario. The scenario-specific model generates a recommended variable fee for the pending transaction based on the prepared data. The recommended variable fee is output to an issuer processing the pending transaction. Feedback is received from the issuer regarding the recommended variable fee and the scenario-specific model is trained using the received feedback.