Mutual Learning Machine Learning Models for Fraud Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In electronic commerce, it is challenging to prepare machine learning data for detecting fraudulent or genuineness states due to the diversification of articles and shortening trend cycles, making it difficult to effectively detect fraudulent or inappropriate content using conventional systems.
Innovation Solution
A detecting device employing a second machine learning model for mutual learning with a first model to generate and discriminate article information across multiple modalities, enabling the detection of target states even when necessary data is scarce, using a combination of noise generation and discriminative neural networks for inference.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional machine learning systems are used to detect fraudulent articles, then detection capability can be achieved, but extensive pre-existing machine learning data is required which is difficult to prepare due to article diversification and shortening trend cycles
Solution Approach 1:
The system performs preliminary action by generating synthetic fraudulent article data before actual detection is needed. The generative model creates artificial fraudulent article examples in advance, which then serve as training data for the detection model, eliminating the need to wait for actual fraudulent cases to accumulate.
Solution Approach 2:
The system creates copies of genuine article data by transforming them into synthetic fraudulent versions through the generative model. These copied and transformed data samples replicate the characteristics of real fraudulent articles without requiring actual fraudulent cases, enabling training of detection models with sufficient data volume.
2Measurement precision
If machine learning data is prepared using actual fraudulent articles, then accurate detection can be achieved, but the process becomes time-consuming and difficult due to constant article changes and diversification
Solution Approach 1:
Instead of manually collecting and preparing actual fraudulent articles, the system creates synthetic copies through the generative model. This copying process automatically generates detection data that captures fraudulent patterns, dramatically reducing the time and effort required for data preparation while maintaining detection accuracy.
Solution Approach 2:
The system performs preliminary data generation by creating synthetic fraudulent examples in advance through automated processes. This preliminary action eliminates the need for time-consuming manual data collection and preparation, allowing the detection model to be trained quickly on freshly generated data that reflects current fraudulent patterns.
3Reliability
If traditional detection systems are used, then existing fraudulent patterns can be identified, but the system lacks adaptability to new and evolving fraudulent methods in dynamic market conditions
Solution Approach 1:
The system implements dynamics by enabling continuous regeneration of detection data through the generative model. As fraudulent methods evolve, the model can dynamically generate new synthetic examples reflecting current patterns, allowing the detection system to adapt continuously without requiring complete retraining on actual fraudulent cases.
Solution Approach 2:
The system performs preliminary generation of adaptive detection data by anticipating the need for new fraudulent patterns. The generative model can proactively create synthetic examples of emerging fraudulent methods based on learned patterns, enabling the detection system to adapt to new threats before they become widespread.
Data Source
AI summary
Disclosed herein is a detecting device including a second machine learning model provided for mutual learning together with a first machine learning model, the first machine learning model being subjected to machine learning so as to generate genuine or fraudulent article information having a plurality of modalities, and the second machine learning model being subjected to machine learning so as to discriminate whether article information having a plurality of modalities is genuine or not, an obtaining section configured to obtain article information having a plurality of modalities, and an estimating section configured to estimate whether the article information that is obtained by the obtaining section and has the plurality of modalities is genuine or not, by using the second machine learning model.