Deep Learning Fraud Detection Using Time Difference Features
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for fraudulent transaction identification, such as logistic regression and random forest, fail to effectively utilize time duration between user operations, leading to incomplete feature extraction and reduced accuracy in detecting fraudulent transactions.
Innovation Solution
A deep learning network is trained using a combination of operation sequences and time difference information to predict fraudulent transactions, employing techniques like LSTM for sequence and time modeling, which enhances the identification of fraudulent patterns by considering temporal associations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional models (logistic regression, random forest) are used for fraudulent transaction identification, then the model structure is simple and easy to implement, but the identification accuracy is insufficient due to failure to utilize time duration between user operations
Solution Approach 1:
The patent introduces time dimension to the traditional transaction identification model by extracting time difference features between adjacent operations. This transforms the data from static operation sequences to dynamic temporal sequences, enabling the model to capture temporal patterns and fraud behaviors that occur over time, thereby improving identification accuracy without fundamentally changing the model architecture
Solution Approach 2:
The patent segments the transaction operation sequence into discrete operations and extracts time difference features between adjacent operations. By dividing the continuous transaction flow into segmented operations with measurable time intervals, the model can analyze temporal patterns and identify fraudulent behaviors that exploit time-based patterns
2Reliability
If time difference information is added to the feature set, then the temporal patterns and fraud behaviors can be captured, but the feature extraction complexity increases
Solution Approach 1:
The patent performs preliminary feature extraction and transformation on the operation sequences and time difference information before feeding them into the classification model. By pre-processing the data to extract meaningful temporal features and prepare them in the required format, the model can focus on classification rather than feature engineering, reducing overall system complexity while maintaining high detection reliability
Data Source
AI summary
Techniques for identifying fraudulent transactions are described. In one example method, an operation sequence and time difference information associated with a transaction are identified by a server. A probability that the transaction is a fraudulent transaction is predicted based on a result provided by a deep learning network, where the deep learning network is trained to predict fraudulent transactions based on operation sequences and time differences associated with a plurality of transaction samples, and where the deep learning network provides the result in response to input including the operation sequence and the time difference information associated with the transaction.