ML-Based Transaction Attribute Allocation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Loyalty programs face challenges in managing reward distribution, particularly on large data sets, as existing methods are prone to misuse, require significant computing resources, and result in latency and inaccuracy, limiting adaptability to customer behavior changes.
Innovation Solution
A processor-implemented method using machine learning models to dynamically allocate transaction-bound attributes by generating a personalized waiting period, reducing the need for manual interventions and scheduled jobs, and optimizing storage and computational resources.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If scheduled job systems are used to transition reward points from pending to available status, then reward points can be made available after return period, but computing resources are significantly required and system complexity increases
Solution Approach 1:
The patent replaces the mechanical scheduled job system with a machine learning-based predictive system. Instead of using time-based scheduled tasks to transition points from pending to available status, the system uses ML models to predict fraudulent transactions and directly determine point availability in real-time, eliminating the need for complex scheduled job management and reducing system complexity while maintaining reliability
Solution Approach 2:
The system enables self-service by allowing the machine learning model to automatically assess transaction risk and determine reward points availability without manual intervention or complex scheduled job coordination. The ML model independently evaluates each transaction and makes real-time decisions about point allocation, simplifying the system architecture
2Reliability
If scheduled job systems are used to monitor and manage reward points, then points can be secured against misuse, but computing resources and continuous monitoring requirements increase operational costs
Solution Approach 1:
The patent applies preliminary action by using the machine learning model to predict fraudulent transactions before they occur. The system pre-assesses transaction risk and determines reward points availability in advance, eliminating the need for continuous monitoring and reducing computing resource consumption during operation while maintaining strong fraud prevention capabilities
Solution Approach 2:
The system transitions from static scheduled job monitoring to dynamic real-time assessment. The machine learning model dynamically evaluates each transaction based on current data and patterns, adjusting reward points availability decisions in real-time without requiring continuous resource-intensive monitoring, thus reducing operational costs while maintaining security
3Measurement precision
If transaction history-based discrimination is used to calculate available reward points, then fraudulent transactions can be identified, but latency increases proportionally to historical transaction volume
Solution Approach 1:
The patent changes the parameters of fraud detection by shifting from analyzing entire historical transaction volumes to using machine learning models that process condensed feature representations. The ML models are trained on historical data but make predictions based on extracted features rather than raw transaction volumes, dramatically reducing processing latency while maintaining detection accuracy
4Reliability
If manual determination of escrow period and member status is performed, then reward distribution can be controlled, but time-intensive operations and human inaccuracies increase
Solution Approach 1:
The patent replaces manual determination processes with automated machine learning-based decision-making. The ML models automatically assess member status, determine appropriate escrow periods, and control reward distribution based on learned patterns from historical data, eliminating human inaccuracies and significantly improving processing efficiency while maintaining reliable control over reward distribution
Data Source
AI summary
A method for dynamically allocating transaction-bound attributes associated with at least one member account by generating a personalized waiting period of the transaction-bound attributes using a machine learning model is provided. The method includes (i) performing a first machine learning (ML) model on a historical transaction dataset, a historical member dataset, and a historical entity dataset, (ii) training a second machine learning model based on correlation and patterns in (a) the predicted transaction dataset, (b) the predicted member dataset, and (c) the predicted entity dataset, (iii) dynamically generating the personalized waiting period for the transaction-bound attributes using the second machine learning model, (iv) dynamically determining, using the second machine learning model, the personalized waiting period for the transaction-bound attributes that are placed in the pending state, and (v) dynamically allocating the transaction-bound attributes associated with at least one member account based on the generated personalized waiting period of the transaction-bound attributes.