Fraudulent Listing Risk Assessment via ML Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Fraudulent listings in online marketplaces pose a significant problem for both marketplaces and payment providers, leading to financial losses and damage to reputation, as they can overwhelm legitimate listings and reduce marketplace usability.
Innovation Solution
A computing system is implemented to conduct risk analysis on listings using a classification algorithm trained with data from confirmed fraudulent and legitimate listings, which scrapes and structures data from marketplaces, transforms it into variables, and classifies new listings based on these features to determine their legitimacy, enabling remediation actions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If risk analysis is conducted on listings, then fraudulent listings can be identified and filtered out, but the system complexity increases due to data scraping, transformation, and classification algorithms
Solution Approach 1:
The system divides the risk analysis process into distinct segments: data scraping module, data transformation module, classification algorithm module, and risk assessment module. Each segment handles a specific aspect of the analysis, making the complex system more manageable and maintainable while improving reliability through specialized processing at each stage.
Solution Approach 2:
The patent introduces intermediary components including a data scraping service that extracts information from marketplaces, a data transformation service that converts raw data into usable features, and a classification service that applies machine learning models. These intermediaries simplify the overall system architecture by encapsulating complex operations in separate, reusable services.
2Measurement precision
If a classification algorithm is trained with scraped data, then the accuracy of fraudulent listing identification improves, but the time required for data collection and processing increases
Solution Approach 1:
The system performs preliminary data scraping and transformation operations continuously in the background, maintaining a preprocessed dataset of listing features ready for classification. This preliminary action reduces the time required when new listings need to be assessed, as the heavy data processing has already been completed beforehand.
Solution Approach 2:
The data scraping and transformation processes operate continuously to maintain an up-to-date database of listing characteristics. This continuous operation ensures that the classification algorithm always has access to current, preprocessed data, eliminating delays associated with batch processing and improving both accuracy and speed.
3Ease of operation
If fraudulent listings are filtered out, then marketplace usability is improved by reducing spam, but legitimate listings may be incorrectly flagged reducing marketplace activity
Solution Approach 1:
The system incorporates feedback mechanisms where classification results are continuously monitored and adjusted. When the classification algorithm flags a listing, the system can review the decision and provide feedback to refine the model. This feedback loop reduces false positives and ensures that legitimate listings are not incorrectly filtered out, maintaining marketplace activity while improving usability.
Solution Approach 2:
The classification algorithm dynamically adjusts its risk thresholds and classification parameters based on marketplace conditions and learning from new data. By changing parameters such as risk tolerance levels and classification cutoffs, the system can adapt to reduce false positives and maintain healthy marketplace activity while still effectively filtering fraudulent listings.
Data Source
AI summary
A system and method for assessing the risk of a listing that transforms information from the listing into variables suitable for a classifier trained to score the riskiness of listings and using the score in addition to predetermined variable constraints to determine whether a listing is fraudulent.