Transaction Overspend Detection via Predictive Distribution Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current online systems lack an effective method to identify unexpectedly high transaction overspend, which can lead to loss and fraud, due to variations in costs caused by factors like taxes, price changes, and item discrepancies.
Innovation Solution
The system employs machine learning models to predict overspend and error terms, generating a distribution for expected overspend, and flags transactions exceeding a predetermined threshold percentile, thereby allowing or declining transactions based on predicted overspend likelihood.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the system allows all transactions with expected cost variations, then transaction throughput is maintained, but the system becomes vulnerable to fraud and unexpected losses
Solution Approach 1:
The system performs preliminary analysis by training machine learning models to predict expected overspend distributions before transactions occur. During transaction processing, the system compares actual costs against pre-established predicted distributions to quickly identify anomalies, maintaining high transaction throughput while detecting fraud accurately
Solution Approach 2:
The system replaces manual fraud review processes with automated machine learning models that predict overspend distributions. The models use historical data to establish expected cost variations and automatically flag transactions that deviate significantly, eliminating the need for human reviewers to examine every transaction while maintaining high detection accuracy
2Reliability
If the system implements strict overspend thresholds, then fraud prevention improves, but legitimate transactions with normal cost variations are incorrectly declined
Solution Approach 1:
The system dynamically adjusts overspend thresholds based on predicted distributions from machine learning models rather than using fixed thresholds. The models calculate expected mean and standard deviation of overspend for each transaction type, allowing the system to adapt thresholds to normal cost variations while maintaining sensitivity to actual fraud patterns
Solution Approach 2:
The system continuously learns from transaction outcomes by incorporating actual overspend data back into the machine learning models. This feedback loop refines the predicted distributions over time, improving the accuracy of dynamic thresholds and reducing false positives while maintaining effective fraud prevention
Data Source
AI summary
An online system receives a request to confirm a transaction that is associated with an order. The system accepts or declines the transaction based on whether an amount associated with the pending transaction is likely to exceed an expected amount of the order by more than a threshold value. To determine the threshold, the system trains a first model to predict an overspend for an order and then trains a second model to predict an amount of error associated with the predictions from the first model. The outputs of the first model and the second model provide a mean and a variance for an expected distribution of the overspend. If the actual overspend amount for the transaction exists in too high of a percentile of the distribution, the transaction may be flagged for review or declined.