GAN-Based Return Validation for Fraud Detection

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current return policies are ineffective in distinguishing between genuine and non-genuine returns, leading to retailers absorbing defective items and incurring losses due to fraudulent return practices.

Innovation Solution

A system utilizing a Generative Adversarial Network (GAN) trained to detect non-genuine returns, which processes customer transaction history and return policy rules to determine the validity of return requests and recommend appropriate processing actions.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Ease of operation

If traditional return policies are used to allow returns within certain conditions, then customer convenience is improved, but loss from fraudulent returns increases

Engineering Contradiction:

Improvecustomer convenienceVSAvoidfinancial loss from fraudulent returns

Core Design Contradiction:

Ease of operationVSLoss of energy

Solution Approach 1:

The patent replaces manual return validation processes with an automated machine learning system. The GAN-based validation engine automatically analyzes return requests, customer behavior patterns, and transaction data to detect fraudulent returns, substituting human judgment with an automated intelligent system that can process returns at scale while maintaining security.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The system implements continuous feedback loops where return validation outcomes, customer responses, and fraud patterns are fed back into the GAN model for ongoing training and improvement. This allows the system to learn from new fraud techniques and refine its detection accuracy over time while adapting to legitimate customer needs.

Inventive Principle:

Principle #23Feedback

2Reliability

If return policy conditions are tightened to prevent fraud, then loss prevention is improved, but customer satisfaction deteriorates

Engineering Contradiction:

Improveloss preventionVSAvoidcustomer satisfaction

Core Design Contradiction:

ReliabilityVSEase of operation

Solution Approach 1:

The patent applies differentiated validation scrutiny to different return requests based on individual customer risk profiles. Legitimate customers with good histories experience smooth, minimal-friction returns, while suspicious patterns trigger enhanced validation. This localized approach ensures fraud prevention measures are applied precisely where needed rather than uniformly to all customers.

Inventive Principle:

Principle #3Local quality

Solution Approach 2:

The system dynamically adjusts validation parameters and decision thresholds based on learned patterns from the GAN model. Rather than using fixed rigid rules, the validation criteria adapt to changing fraud patterns and customer behaviors, allowing the system to maintain high reliability while accommodating legitimate return variations.

Inventive Principle:

Principle #35Parameter changes

3Measurement precision

If manual review of return requests is performed to detect fraud, then detection accuracy is improved, but processing time increases

Engineering Contradiction:

Improvefraud detection accuracyVSAvoidreturn processing time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The patent replaces time-consuming manual review processes with automated GAN-based validation that operates in real-time. The machine learning system analyzes customer behavior patterns, transaction histories, and return request details instantaneously, providing fraud detection accuracy comparable to or exceeding manual review while eliminating processing delays.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The system performs preliminary analysis of customer return patterns and risk profiles before actual return requests are submitted. By pre-establishing risk assessments and validation criteria based on historical data, the system is prepared to make rapid real-time decisions when returns are requested, eliminating the need for time-consuming ad-hoc manual reviews.

Inventive Principle:

Principle #10Preliminary action

4Measurement precision

If GAN-based validation system is implemented to detect non-genuine returns, then fraud detection capability is improved, but system complexity increases

Engineering Contradiction:

Improvefraud detection capabilityVSAvoidsystem complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The patent introduces a specialized GAN-based validation engine as an intermediary component between return requests and processing decisions. This dedicated fraud detection module encapsulates the complex machine learning logic in a self-contained system that interfaces with existing return management infrastructure through standardized protocols, isolating complexity to a specific component rather than distributing it throughout the entire system.

Inventive Principle:

Principle #24Intermediary (Mediator)

Data Source

PatentUS11830011B2Dynamic return optimization for loss prevention based on customer return patterns

Publication Date: 2023.11.28 INTERNATIONAL BUSINESS MACHINE CORPORATION

AI summary

Approaches presented herein enable dynamically determining a validity of a return. More specifically, a system obtains a return request from a customer, a transaction history of the customer, and a set of return policy rules. A generative adversarial network (GAN) trained to detect non-genuine returns is applied to the return request. The GAN uses, among other this, the transaction history of the customer and the set of return policy rules as parameters of the GAN. Based on an output of the GAN, at least one return processing action is recommended and implemented.