ML Transaction Classification for Subjective Approval Decisions

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing automated systems lack the ability to make subjective decisions beyond pre-approved categories, leading to inaccuracies in transaction approval and denial, and require manual input for transactions that do not fit within these categories.

Innovation Solution

A machine learning system using pattern recognition to analyze transaction data and metadata, applying dynamic rules to classify transactions as healthy, moderate, or critical, and automate approval or denial without manual intervention.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If manual review is used for transaction classification, then accuracy in handling new and unforeseen transactions is improved, but processing time and operational complexity increase significantly

Engineering Contradiction:

Improvetransaction classification accuracyVSAvoidprocessing time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The system performs preliminary classification of transactions using machine learning models before human review. The ML system pre-processes and categorizes transactions, flagging only those requiring manual attention, thereby reducing overall processing time while maintaining accuracy for complex cases

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

A machine learning intermediary system is introduced between the transaction input and final approval. This intermediary automatically classifies transactions and provides recommendations to human reviewers, reducing their workload while maintaining high accuracy through the combined ML-human decision-making process

Inventive Principle:

Principle #24Intermediary (Mediator)

2Productivity

If pre-approved categories are used for transaction classification, then processing speed is improved, but adaptability to new and unforeseen transactions deteriorates

Engineering Contradiction:

Improvetransaction processing speedVSAvoidhandling of new transaction types

Core Design Contradiction:

ProductivityVSAdaptability or versatility

Solution Approach 1:

The classification system transitions from static pre-approved categories to dynamic machine learning models that continuously adapt. The ML models are trained on historical data and can evolve to recognize new transaction patterns, maintaining both speed through automation and adaptability through continuous learning

Inventive Principle:

Principle #15Dynamics

Solution Approach 2:

The system changes the parameters of classification from fixed categorical rules to flexible probabilistic models. By using ML algorithms that can adjust their decision boundaries based on learned patterns, the system maintains high processing speed while becoming adaptable to new transaction types through parameter optimization rather than rule updates

Inventive Principle:

Principle #35Parameter changes

3Productivity

If automated systems are used for transaction approval, then operational efficiency is improved, but reliability in making subjective decisions beyond pre-approved categories deteriorates

Engineering Contradiction:

Improveoperational efficiencyVSAvoiddecision accuracy for complex transactions

Core Design Contradiction:

ProductivityVSReliability

Solution Approach 1:

The system implements feedback loops where transaction outcomes are continuously fed back into the ML training process. This allows the automated system to learn from past decisions and improve its reliability over time, while maintaining high operational efficiency through automation of the decision-making process

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

The ML system performs self-training and self-improvement by automatically learning from historical transaction data and outcomes. This self-service capability enables the system to enhance its own reliability for complex subjective decisions without requiring constant manual reconfiguration, maintaining operational efficiency while improving decision accuracy

Inventive Principle:

Principle #25Self-service

Data Source

PatentUS20260065280A1Machine learning system for automated subjective decision management

Publication Date: 2026.03.05 AMERICAN EXPRESS (INDIA) PTE LTD

AI summary

Disclosed herein are system, method, and computer program product embodiments for a transaction management system. The transaction management system may receive a request, e.g., transaction data, from a merchant system. The transaction management system may also receive metadata associated with a corresponding transaction card holder. The transaction management system may analyze the transaction using a system employing machine learning model(s) to perform functions such as pattern recognition and anomaly detection. Transaction management system may also use machine learning systems to manage approval and payment for transactions, especially in the context of new and unforeseen transactions. Transactions may be classified so that the transaction may be automatically approved, in need of additional review, or denied. Each of these classifications may be decided without input or oversight from users.