Auto AI Model Explainability via Business Rule Correlation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current MLOps and auto AI tools prioritize model performance metrics like accuracy over maintainability of model explainability, failing to ensure consistent and credible results across model refreshes, especially in regulated industries where transparency is crucial.
Innovation Solution
A computer-implemented method for building prediction models that integrates data modeling quality metrics with business rule correlations to generate AI models that meet both accuracy and explainability requirements, allowing for early stopping and embedding explainability into the model generation process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If current MLOps and auto AI tools focus on optimizing model performance metrics like accuracy, then model prediction accuracy is improved, but model explainability deteriorates across model refreshes
Solution Approach 1:
The patent combines model performance optimization with explainability maintenance by integrating business rule validation into the model training and refresh process. The system simultaneously optimizes for accuracy metrics while validating that model predictions continue to satisfy predefined business rules, thereby merging performance and explainability objectives into a unified framework.
Solution Approach 2:
The patent implements a feedback mechanism where business rule validation results are used to guide model refresh decisions. When a retrained model fails to meet explainability standards (business rule violations), the system provides feedback to prevent deployment and triggers further training adjustments, creating a closed-loop system that maintains both accuracy and explainability.
2Adaptability or versatility
If models are retrained periodically as data distribution changes, then model adaptability is improved, but explainability consistency deteriorates
Solution Approach 1:
The patent applies preliminary action by establishing business rules and explainability criteria before model training begins. These predefined rules serve as constraints that guide the model training process, ensuring that adaptability improvements through retraining do not compromise explainability consistency. The business rules are set in advance to prevent explainability degradation during model refreshes.
Solution Approach 2:
The patent utilizes parameter changes by adjusting model training parameters and selection criteria based on business rule compliance. When retraining models to adapt to changing data distributions, the system modifies training parameters or selects different model configurations that maintain adherence to business rules, thereby preserving explainability consistency while achieving adaptability.
3Reliability
If qualitative explainability aspects are quantized and linked to business rules, then model credibility is improved, but system complexity increases
Solution Approach 1:
The patent extracts explainability validation as a separate, modular component that operates independently from the core model training process. By taking out the business rule validation logic as a distinct layer, the system can quantify and assess explainability without fundamentally complicating the model training architecture, thereby improving credibility while managing system complexity through modular design.
Solution Approach 2:
The patent introduces business rules as an intermediary layer between the model and the explanation requirements. Instead of directly embedding complex explainability mechanisms into the model, the system uses business rules as mediators that translate qualitative explainability aspects into quantifiable validation criteria, simplifying the overall system architecture while maintaining model credibility.
Data Source
AI summary
A computer-implemented method for model building with explainability is provided. The method includes receiving, by a hardware processor, a first metric and a second metric of minimum model performance. The first metric relates to data modeling quality and the second metric relates to model to business rule correlations. The method further includes performing, by the hardware processor, auto Artificial Intelligence model generation responsive to training data and a combination of the first and the second metrics of minimum model performance to obtain a model that is trained and meets model prediction accuracy and model prediction explainability requirements represented by the combination of the first and the second metrics of minimum model performance.