Machine Learning Error Detection Using Explainable Feature Importance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing machine learning models struggle to provide actionable insights into when and how they have erred, making it challenging to correct erroneous predictions.
Innovation Solution
A system and method that determine global-level importance magnitude and direction labels for explainable features of a machine learning model, allowing for the identification and correction of erroneous predictions through local and global-level importance analysis.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If machine learning models perform predictions based on data patterns, then prediction capability is improved, but ability to provide actionable insights into errors deteriorates
Solution Approach 1:
The patent segments the model's decision-making process into interpretable components by identifying and ranking explainable features. This segmentation allows the system to break down complex predictions into individual feature contributions, making error analysis possible without reducing overall prediction capability.
Solution Approach 2:
The patent introduces an intermediary explanation layer that mediates between the black-box model predictions and user understanding. This intermediary component generates human-interpretable feature importance rankings that bridge the gap between model operations and actionable error insights.
2Measurement precision
If machine learning models use complex algorithms, then prediction accuracy is improved, but interpretability of predictions deteriorates
Solution Approach 1:
The patent introduces an intermediary explanation layer that mediates between the black-box model predictions and user understanding. This intermediary component generates human-interpretable feature importance rankings that bridge the gap between model operations and actionable error insights.
Solution Approach 2:
The patent employs visual indicators (such as colored bars or highlights) to represent feature importance magnitudes and directions. This visual encoding transforms complex numerical importance values into intuitive visual representations that are easy to interpret while preserving prediction accuracy.
3Productivity
If machine learning models operate as black-box systems, then computational efficiency is improved, but detection of erroneous predictions deteriorates
Solution Approach 1:
The patent enables the model to self-diagnose errors by automatically generating feature importance rankings for its own predictions. This self-service mechanism allows the system to detect erroneous predictions without external intervention, maintaining computational efficiency while improving error detection capability.
Solution Approach 2:
The patent implements a feedback loop where feature importance rankings are generated for each prediction and used to detect errors. This feedback mechanism provides actionable insights into model errors while maintaining the computational efficiency of black-box operations by using the same model infrastructure.
Data Source
AI summary
A system includes a memory having instructions therein and at least one processor in communication with the memory. The at least one processor is configured to execute the instructions to determine a global-level importance magnitude value for a global-level importance of an explainable feature of a machine learning base model based on a first prediction of the machine learning base model. The at least one processor is also configured to execute the instructions to determine a global-level importance direction label for the global-level importance of the explainable feature based on the first prediction. The at least one processor is also configured to execute the instructions to generate a communication for presentation to a user based on a second prediction of the machine learning base model, based on the global-level importance magnitude value, and based on the global-level importance direction label.