Automated Feature Selection for Real-Time Fraud Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Machine learning algorithms used for fraud detection in real-time transactions are hindered by the complexity of large feature sets, making it time- and resource-consuming to determine whether a transaction is authorized or unauthorized within a short time frame.
Innovation Solution
A system and method that narrows down an initial set of hundreds or thousands of machine learning features to a smaller set of 5-10 features using algorithms like the Markov blanket algorithm, constructs a decision tree, and generates rules with high accuracy, allowing for real-time transaction classification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a large number of machine learning features (hundreds or thousands) are used to identify future bad transactions, then the accuracy of fraud detection is improved, but the time and computing resources required increase significantly, making real-time determination impossible
Solution Approach 1:
The patent extracts and removes redundant features from the large initial feature set through automated machine learning processes. The system identifies and eliminates features that do not contribute significantly to fraud detection accuracy, retaining only the essential subset of features needed for real-time classification. This extraction process reduces the feature set from hundreds or thousands to a manageable number while preserving detection accuracy.
Solution Approach 2:
The patent segments the large feature set into multiple subsets through automated feature selection processes. Different subsets of features are evaluated and combined in optimized ways to achieve accurate fraud detection. This segmentation allows the system to process features in manageable groups rather than dealing with the entire large set simultaneously, reducing computational burden.
2Measurement precision
If a large number of machine learning features are used to identify future bad transactions, then the accuracy of fraud detection is improved, but the computing resource burden increases, making the system impractical for real-time operation
Solution Approach 1:
The system extracts and eliminates computationally redundant features from the large feature set. By identifying and removing features that provide minimal additional value to detection accuracy, the system significantly reduces the computing resources required for real-time transaction classification while maintaining fraud detection effectiveness.
Solution Approach 2:
The patent changes the parameters of the feature set by transforming and optimizing feature representations. Automated machine learning processes adjust feature selections and combinations to achieve the optimal balance between detection accuracy and computational efficiency, making the system practical for real-time operation with available computing resources.
3Reliability
If hundreds or thousands of features are manually selected and processed, then comprehensive fraud detection coverage is achieved, but the process becomes time-consuming and resource-intensive
Solution Approach 1:
The patent implements self-service automated machine learning processes that automatically perform feature selection, evaluation, and optimization without requiring manual intervention. The system autonomously identifies the most effective feature subsets for fraud detection, eliminating the need for time-consuming manual feature processing while maintaining comprehensive fraud detection coverage.
Solution Approach 2:
The system automatically adjusts and optimizes feature selection parameters through machine learning algorithms. This automated parameter optimization enables the system to achieve comprehensive fraud detection coverage efficiently, transforming the manual, resource-intensive feature processing into an automated, high-productivity process.
Data Source
AI summary
Systems and methods for machine learning feature processing are disclosed. An example method includes: selecting a predefined number of machine-based transactions; identifying a first set of features based on the predefined number of machine-based transactions; and applying a machine learning feature selection algorithm to identify a second set of features. The second set of features is smaller than the first set of features, and the second set of features includes no more than a predefined number of features. The method also includes constructing a decision tree based on the second set of features, wherein the decision tree has a depth that is less than a predefined depth. The method optionally includes limiting the total number of features in the second set to no greater than the predefined number of features.