Segmented Decision Thresholds Using Linear Programming
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current computer systems evaluating transaction requests using machine learning algorithms face performance degradation due to incorrectly rejecting or granting transactions, leading to inefficiencies in resource utilization and potential security or financial losses, as they often use a single decision threshold score for all transactions without accounting for segment-specific variations.
Innovation Solution
Implementing a computer system that calculates segment-level decision threshold scores using linear programming to optimize transaction evaluation, allowing for different thresholds based on constraints such as decline rate, catch rate, and false positive rate for various segments of transaction data, thereby improving the accuracy and efficiency of transaction request evaluations without modifying the machine learning algorithm.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a single decision threshold score is used for all transactions, then the system operation is simple, but the transaction evaluation accuracy deteriorates due to segment-specific variations being ignored
Solution Approach 1:
The patent divides the transaction dataset into multiple segments based on characteristics such as transaction type, user behavior patterns, or risk categories. Each segment receives its own customized decision threshold score calculated through linear programming, allowing the system to account for segment-specific variations and improve evaluation accuracy without overwhelming complexity.
Solution Approach 2:
The patent implements different decision threshold scores for different transaction segments, where each segment has optimized thresholds tailored to its specific characteristics. This local customization of evaluation criteria improves accuracy for each segment while the overall system remains manageable through automated calculation methods.
2Measurement precision
If segment-level decision threshold scores are calculated using linear programming, then the transaction evaluation accuracy is improved, but the computational complexity increases
Solution Approach 1:
The patent pre-calculates decision threshold scores for multiple segments using linear programming during an offline or batch processing phase. By performing these complex calculations in advance rather than in real-time for each transaction, the system achieves high accuracy thresholds while keeping real-time operational complexity manageable.
Solution Approach 2:
The patent creates a simplified copy or lookup table of pre-calculated decision threshold scores for each segment. During transaction evaluation, the system simply retrieves the appropriate pre-computed threshold based on the transaction's segment classification, avoiding the need to perform complex linear programming calculations for every individual transaction.
3Productivity
If uniform decision thresholds are applied to all segments, then the system operation is efficient, but the correct granting or rejection of transactions deteriorates due to segment-specific variations
Solution Approach 1:
The patent segments transactions into distinct categories and applies customized decision thresholds to each segment. This segmentation allows the system to maintain high throughput by using efficient lookup-based evaluation while simultaneously improving decision correctness by accounting for segment-specific characteristics through optimized thresholds.
Solution Approach 2:
The patent changes the decision threshold parameter dynamically based on the transaction segment. By adjusting this critical parameter according to segment characteristics, the system achieves both efficiency (through automated parameter selection) and reliability (through segment-optimized parameters) without requiring complex real-time calculations.
4Measurement precision
If multiple segment-level thresholds are implemented, then the transaction evaluation accuracy is improved, but the device complexity increases
Solution Approach 1:
The patent pre-computes and stores multiple segment-level decision thresholds using linear programming optimization. This preliminary action consolidates the complexity of managing multiple thresholds into an offline calculation phase, allowing the online evaluation system to simply retrieve and apply the appropriate threshold based on segment classification, thereby maintaining structural simplicity during operation.
Data Source
AI summary
A computer system is configured to receive a dataset that includes a plurality of transaction request records and is divisible into a plurality of segments. Each transaction request record includes an individual score calculated by a machine learning algorithm. The computer system also receives a plurality of constraints. The computer system is configured to calculate, using a linear programming algorithm, a decision threshold score for a particular segment of the plurality of segments using the transaction request records. The computer system is configured to provides access to the calculated decision threshold score to a production computer system. The production computer system is configured to use the decision threshold score to evaluate a subsequent transaction request corresponding to the particular segment.