Miscoded Tag Detection Using Risk-Scored Record Permutations
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing anomaly detection systems in financial journal entries, such as those for accounting and auditing, are labor-intensive and require constant maintenance due to their reliance on brittle business rules, failing to identify complex patterns and adapt to changing business needs.
Innovation Solution
A two-stage machine learning solution using a machine learning scorer and an unsupervised Bayesian network model to identify anomalies, iteratively remove fields, and generate permutations to replace anomalous values, thereby automating the detection and correction process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If rules-based systems are used for anomaly detection in journal entries, then mechanical errors can be caught, but the systems require constant maintenance and cannot identify complex patterns
Solution Approach 1:
The patent replaces the mechanical rules-based system with a machine learning model that automatically learns patterns from historical data. The model uses supervised learning to train on labeled anomalies and unsupervised learning to detect novel patterns, eliminating the need for manual rule creation and maintenance while improving adaptability to changing business conditions.
Solution Approach 2:
The system implements dynamic adaptation by continuously retraining the machine learning model on new data and allowing it to evolve with changing business patterns. The model can identify both known and emerging anomaly patterns without requiring explicit rule updates, making the system flexible and responsive to business changes.
2Productivity
If rules-based systems are deployed to detect anomalies, then initial error detection works, but maintenance burden increases over time
Solution Approach 1:
The machine learning model performs self-learning and self-updating by automatically training on new data and adapting to changing patterns. This eliminates the need for manual rule maintenance and reduces the time investment required over time, as the system improves autonomously rather than requiring continuous human intervention.
Solution Approach 2:
The system performs preliminary training on historical data to establish baseline patterns before deployment. Once trained, the model is ready to detect anomalies without requiring ongoing rule creation or adjustment, significantly reducing maintenance time compared to rules-based systems that must be continuously updated.
3Measurement precision
If traditional accounting review processes are used, then thorough examination is possible, but labor-intensive processes require many professionals
Solution Approach 1:
The machine learning model serves multiple functions: it detects various types of anomalies (misclassification, duplicate entries, fraudulent transactions), adapts to different business contexts, and can be deployed across multiple organizations. This single system replaces the need for multiple specialized reviewers and complex organizational structures.
Solution Approach 2:
The patent replaces the mechanical process of human review with an automated machine learning system that can examine data at scale with consistent thoroughness. The model analyzes patterns across entire datasets without fatigue or variation, eliminating the need for large teams of professionals while maintaining or improving review quality.
Data Source
AI summary
The techniques described herein relate to a method including: receiving, by a processor, a data record having a plurality of fields; generating, by the processor, a risk score for the data record using a predictive model; determining, by the processor, that the data record is a potential anomaly based on the risk score; identifying, by the processor, an anomalous field from the plurality of fields; generating, by the processor, a plurality of permutations of the data record, the plurality of permutations generated by changing a value of the anomalous field; and outputting, by the processor, a replacement record selected from the plurality of permutations, the replacement record having a field value for the anomalous field that generates a lowest risk score among the plurality of permutations.