RPA AI Model Drift Detection With Automatic Retraining
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing robotic process automation (RPA) technologies lack effective methods for detecting and correcting AI/ML model drift, which can lead to inaccurate predictions and autonomous actions by RPA robots without human supervision.
Innovation Solution
A computer program and method for AI/ML model drift detection and correction in RPA, which analyzes input data and model execution results to determine data and model drift, and triggers retraining of the AI/ML model when drift is detected, ensuring the model meets performance thresholds before deployment.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If AI/ML models are deployed for autonomous RPA tasks without continuous monitoring, then automation efficiency is improved, but prediction accuracy deteriorates over time due to model drift
Solution Approach 1:
The patent implements continuous monitoring of AI/ML model predictions in RPA workflows, comparing actual predictions against expected ranges and triggering alerts or corrections when drift is detected. This feedback mechanism maintains prediction accuracy while preserving automation efficiency by only intervening when necessary.
Solution Approach 2:
The system establishes predetermined acceptable ranges for model predictions during configuration and proactively monitors for drift before it causes significant errors. By detecting drift early and triggering retraining or correction procedures in advance, the system prevents accuracy degradation while maintaining continuous operation.
2Reliability
If continuous monitoring and retraining of AI/ML models is implemented, then prediction accuracy is maintained, but system complexity increases
Solution Approach 1:
The patent enables AI/ML models to self-monitor their own prediction accuracy and self-trigger retraining procedures when drift is detected. The system automatically compares predictions against expected ranges, identifies drift conditions, and initiates corrective actions without requiring complex external monitoring infrastructure.
Solution Approach 2:
The system monitors changes in prediction parameters and statistical distributions to detect drift. By tracking parameter shifts rather than implementing complex behavioral analysis, the system maintains prediction accuracy through simpler, more manageable monitoring mechanisms.
3Difficulty of detecting and measuring
If statistical analysis of predictions is performed continuously, then drift detection capability is improved, but processing time increases
Solution Approach 1:
The patent implements selective statistical analysis that performs comprehensive drift detection only when necessary, rather than continuously analyzing all predictions at full depth. The system uses efficient statistical methods and triggers detailed analysis only when preliminary indicators suggest potential drift, reducing processing overhead while maintaining detection capability.
Solution Approach 2:
The system performs statistical drift detection at periodic intervals rather than continuously analyzing every prediction in real-time. This periodic approach maintains adequate drift detection capability while significantly reducing processing time and computational resource requirements compared to continuous analysis.
Data Source
AI summary
Artificial intelligence (AI)/machine learning (ML) model drift detection and correction for robotic process automation (RPA) is disclosed. Information is analyzed pertaining to input data for an AI/ML model to determine whether data drift has occurred, analyze information pertaining to results from execution of the AI/ML model to determine whether model drift has occurred, or both. When, based on the analysis of the information, a change condition is found, a change threshold is met or exceeded, or both, the AI/ML model is retrained. The retrained AI/ML model may then be deployed to provide better predictions on real world data.