Transaction Profile Tagging for Fraud Detection

Solution Overview

Problem

Current systems consume unnecessary computational and network resources when identifying fraudulent transactions, as they often process transactions without sufficient inquiry, leading to false positives and unnecessary resource allocation.

Innovation Solution

The method involves generating transaction profile tags by processing recorded transactions with macro- and micro-encoders to cluster features into subsets, allowing for the tagging of transaction profiles, which enables more efficient identification of suspect transactions by comparing new transactions against profile criteria, reducing false positives and resource consumption.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If predetermined thresholds are used to identify fraudulent transactions, then fraudulent transactions can be detected, but computational and network resources are consumed unnecessarily due to false positives and processing of non-fraudulent transactions

Engineering Contradiction:

Improvefraud detection accuracyVSAvoidcomputational resource consumption

Core Design Contradiction:

ReliabilityVSLoss of energy

Solution Approach 1:

The system performs preliminary actions by generating profile tags from historical transaction data before actual fraud detection occurs. These profile tags pre-characterize account behavior patterns, enabling faster and more accurate real-time fraud detection without requiring extensive computational resources during transaction processing.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The fraud detection system is segmented into multiple components: profile generation module that creates account-specific profiles from historical data, and transaction evaluation module that uses these profiles to assess new transactions. This segmentation allows the system to avoid processing all transactions uniformly, reducing unnecessary computational overhead.

Inventive Principle:

Principle #1Segmentation

2Reliability

If predetermined thresholds are used to identify fraudulent transactions, then fraudulent transactions can be detected, but network resources are consumed unnecessarily due to transmission and processing of non-fraudulent transactions

Engineering Contradiction:

Improvefraud detection accuracyVSAvoidnetwork resource consumption

Core Design Contradiction:

ReliabilityVSLoss of energy

Solution Approach 1:

The system performs preliminary actions by generating profile tags from historical transaction data before actual fraud detection occurs. These profile tags pre-characterize account behavior patterns, enabling faster and more accurate real-time fraud detection without requiring extensive computational resources during transaction processing.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system extracts only the essential profile tags from historical transaction data that are most relevant for fraud detection. By taking out only the critical features and storing them as compact profile tags, the system reduces the amount of data that needs to be transmitted and processed, thereby reducing network resource consumption.

Inventive Principle:

Principle #2Taking out (Extraction)

3Ease of manufacture

If simple threshold-based fraud detection is used, then implementation is straightforward, but false-positive fraud identifications increase leading to unnecessary additional processing

Engineering Contradiction:

Improvesystem implementation simplicityVSAvoidprocessing time for false positives

Core Design Contradiction:

Ease of manufactureVSLoss of time

Solution Approach 1:

The system performs preliminary actions by generating profile tags from historical transaction data before actual fraud detection occurs. These profile tags pre-characterize account behavior patterns, enabling faster and more accurate real-time fraud detection without requiring extensive computational resources during transaction processing.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system changes parameters by moving from simple fixed thresholds to dynamic, account-specific profile tags that capture nuanced transaction patterns. These profile tags include multiple dimensions of account behavior that can be adjusted and refined over time, improving detection accuracy while maintaining implementation feasibility.

Inventive Principle:

Principle #35Parameter changes

Data Source

PatentUS11790013B2Systems and methods for generating transaction profile tags

Publication Date: 2023.10.17 VISA INTERNATIONAL SERVICE ASSOCIATION

AI summary

Methods for generating transaction profile tags from profile transaction activity may include receiving a transaction profile including recorded transactions, associating at least one transaction label with each of the transactions, the labels associated with transaction types, generating a set of profile features based on the recorded transactions from the transaction profile, encoding the set of profile features with a macro-encoder into a first-reduced set, clustering the first-reduced set into at least two subsets, each associated with a macro-profile tag, and tagging the transaction profile with one of the macro-profile tags. Methods may also include encoding the set of profile features with a micro-encoder selected based on the tagged macro-profile tag, clustering the second-reduced set into a plurality of subsets associated with account profile types, respectively, and tagging the transaction profile with a tag associated with the account profile type. Systems and computer program products are also provided.