Rule-Based Counterfactual Explanation for Machine Learning Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Black-box machine learning models lack interpretability, making it difficult to understand and trust their prediction results, especially in critical fields like security, transportation, and healthcare, where existing counterfactual explanation techniques are not rule-based and fail to provide actionable insights for altering undesired predictions.
Innovation Solution
A rule-based counterfactual explanation method that extracts rules from training data, determines matching and counterfactual rules, and performs multi-objective optimization to generate explanations that indicate conditions to alter prediction results, providing actionable advice for changing undesired outcomes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If black-box machine learning models are used to achieve high prediction accuracy, then prediction accuracy is improved, but interpretability deteriorates making it difficult to understand and trust prediction results
Solution Approach 1:
The patent introduces rule-based explanation models as intermediaries between the black-box machine learning model and users. These explanation models generate human-understandable rules that mediate the interpretation of prediction results, allowing users to comprehend the model's decision-making process without sacrificing the high accuracy of the original black-box model
Solution Approach 2:
The patent segments the explanation process into multiple components: extracting rules from training data, determining matching rules for specific samples, generating counterfactual rules, and performing multi-objective optimization. This segmentation allows the complex interpretation task to be broken down into manageable steps that preserve both accuracy and interpretability
2Loss of information
If existing counterfactual explanation techniques are used, then some explanation capability is provided, but they are not rule-based and fail to provide actionable insights for altering undesired predictions
Solution Approach 1:
The patent changes the parameters of counterfactual explanation by introducing rule-based constraints and multi-objective optimization functions that specifically target actionable insights. The optimization process adjusts counterfactual samples to satisfy both fidelity to the original model and adherence to interpretable rules, thereby providing actionable guidance for altering predictions
Solution Approach 2:
The patent implements a feedback mechanism where the rule-based explanation model continuously refines counterfactual explanations based on multi-objective optimization. The system evaluates whether generated counterfactuals provide actionable insights and adjusts the explanation process accordingly, ensuring that users receive practical guidance for changing prediction outcomes
Data Source
AI summary
A method of explaining prediction results of a machine learning model includes: extracting multiple rules based on a training sample set for training the machine learning model and corresponding known labels; determining one or more matching rules to which a sample to be predicted conforms among the rules; generating an explanation model for the machine learning model, wherein the explanation model provides an explanation of a prediction result generated by the machine learning model with respect to a single sample to be predicted; generating counterfactual rules corresponding to the matching rules respectively; determining training samples conforming to one of the counterfactual rules, and forming a counterfactual candidate set including the determined training samples; and performing multi-objective optimization on the counterfactual candidate set to generate a counterfactual explanation. The counterfactual explanation provides conditions to be satisfied by the sample to be predicted in order to alter the prediction result.