Fraud Detection Model Framework Using Serialized Transaction Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems for evaluating electronic transactions face challenges in efficiently generating robust computer models that can adapt to dynamic fraudulent tactics and user population changes, leading to costly and time-consuming processes due to reliance on offline data and inefficient database access.
Innovation Solution
A framework that enables efficient selection, adjustment, and evaluation of candidate features using online transaction data, serializing transaction records for faster access, and providing an API for data scientists to create and test models, allowing for the generation of machine learning models that can work directly with online databases without additional implementation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If computer models are constantly generated to target new fraud tactics, then fraud detection performance is improved, but development cost and time increase
Solution Approach 1:
The system performs preliminary actions by pre-processing and serializing transaction data into a format ready for model generation. Candidate features are pre-computed and stored, so when fraud tactics evolve, models can be rapidly generated without starting from scratch. This preliminary preparation resolves the contradiction by enabling quick model updates (improving reliability) without proportionally increasing complexity.
Solution Approach 2:
The system creates simplified copies of transaction data in serialized format that can be rapidly processed for model generation. Instead of working with complex raw transaction records, the system uses copied and structured feature representations, allowing multiple model iterations without linearly increasing complexity while maintaining detection performance.
2Loss of time
If offline data is used for model generation, then data processing time is reduced, but model accuracy and relevance to current fraud tactics decrease
Solution Approach 1:
The system introduces an intermediary layer - a database of serialized transaction features - that bridges offline data storage and online model generation. The transaction data is pre-processed offline and stored in a serialized format with pre-computed candidate features. This intermediary structure allows rapid online access without reprocessing raw data, enabling both fast model generation and high accuracy using current transaction patterns.
Solution Approach 2:
Data is pre-processed offline into serialized format with candidate features pre-computed and stored in the database. This preliminary action separates the time-consuming processing from the model generation phase, allowing rapid model creation online while maintaining accuracy through access to current transaction data patterns.
3Stability of the object's composition
If complex data structures are used for transaction records, then data organization is improved, but data access speed and manipulation efficiency decrease
Solution Approach 1:
The system extracts essential features from complex transaction data structures and stores them in a simplified serialized format. Instead of navigating complex nested data structures during model generation, the system accesses pre-extracted candidate features directly from the serialized database, maintaining data organization benefits while dramatically improving access speed and manipulation efficiency.
Data Source
AI summary
Methods and systems are presented for providing a framework to generate computer models for detecting fraudulent electronic transactions. The framework selects and retrieves a set of transaction records from an online database system based on one or more criteria. The framework then enriches the set of transaction records with additional data related to the set of transaction records. The additional data of each transaction record may include transaction data of another transaction record selected based on a common attribute shared with the transaction record. The framework serializes the transaction data and the additional data associated with the set of transaction records. A set of features is derived based on analyzing the serialized data. The framework then generates a computer model for detecting fraudulent electronic transaction based on the set of features.