Adaptive Abuse Detection System with Dynamic Threshold Adjustment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current systems for detecting abusive buyer behavior in e-commerce face challenges in efficiently identifying fraudulent refund claims, as they rely on limited bandwidth for case review agents and struggle to optimize risk score thresholds for accurate categorization and restriction of abusive accounts.
Innovation Solution
An adaptive system that uses a machine learning model to calculate risk scores for buyer accounts, coupled with abuse component modules to categorize accounts into specific abuse categories, and adjusts score thresholds based on restriction rates to maximize predictive ability and reduce unnecessary reviews.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the service provider uses machine learning models to compute risk scores for buyer accounts, then the ability to detect abusive behavior is improved, but the bandwidth limitation of case review agents causes bottlenecks in processing flagged accounts
Solution Approach 1:
The patent segments the detection system into multiple abuse component modules (e.g., module 230A, 230B, 230C) that independently evaluate different aspects of buyer behavior. Each module processes accounts in parallel, dividing the overall review workload and increasing throughput while maintaining detection accuracy through specialized evaluation of specific abuse patterns.
Solution Approach 2:
The system dynamically adjusts score thresholds for different abuse component modules based on observed performance metrics and restriction rates. By changing the parameter thresholds adaptively, the system optimizes the balance between detection sensitivity and review volume, ensuring that case review agents receive prioritized accounts with higher abuse likelihood while maintaining overall detection effectiveness.
2Reliability
If the service provider lowers the score threshold to catch more abusive accounts, then detection coverage is improved, but the number of false positives increases causing unnecessary reviews
Solution Approach 1:
The patent divides the detection process into specialized abuse component modules, each targeting specific types of abusive behavior with optimized thresholds. This segmentation allows the system to maintain high detection coverage across different abuse patterns without uniformly lowering thresholds, reducing false positives while catching diverse abusive behaviors through multiple specialized evaluation paths.
Solution Approach 2:
The system implements feedback mechanisms where restriction rates from case review agents are fed back into the threshold adjustment logic. When certain abuse component modules show high restriction rates, their thresholds are adjusted to reduce false positives. This continuous feedback loop maintains detection coverage while optimizing review efficiency by dynamically balancing sensitivity and precision based on actual performance data.
Data Source
AI summary
A system performs operations that include identifying a first subset of accounts from a set of accounts, each account in the first subset of accounts satisfying a first abuse score threshold for a first time period, the first abuse score threshold corresponding to a first buyer abuse component. The operations further include determining a first restriction rate for the first subset of accounts based on a number of accounts in the first subset of accounts that have been restricted for potential abuse. The operations also include comparing the first restriction rate with respective restriction rates of one or more other subsets of the set of accounts that correspond to one or more other abuse components, and based on the comparing, determining whether to adjust the first abuse score threshold.