Prediction Model Segmentation for Interpretable Decision Logic
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Advanced machine learning technologies have resulted in prediction models that are 'black-boxed,' making it difficult to interpret and explain the grounds of prediction results, particularly in medical contexts where healthcare professionals need to understand the rationale behind recommended interventions.
Innovation Solution
A computer system generates a prediction model that balances high predictive performance with the ability to present reproducible prediction logic by using a dataset generated based on an extraction criterion, allowing for the extraction of learning data similar to the prediction target data, and outputs decision logic that explains the prediction process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If advanced machine learning technology is used to generate prediction models, then predictive performance is improved, but the model structure becomes complicated and black-boxed, making interpretation difficult
Solution Approach 1:
The patent segments the prediction model into two distinct components: a black-box prediction model for high predictive performance and a separate decision logic component for interpretability. This segmentation allows each component to fulfill its specific function without compromising the other, resolving the contradiction between predictive performance and interpretability.
Solution Approach 2:
The patent introduces decision logic as an intermediary element that bridges the gap between the black-box prediction model and the user. This intermediary translates the opaque predictions into human-understandable reasoning, allowing users to interpret the model's decisions without sacrificing predictive accuracy.
2Measurement precision
If complex prediction models are used, then predictive accuracy is improved, but the ability to explain prediction grounds deteriorates
Solution Approach 1:
The patent separates the prediction function from the explanation function. The black-box model handles prediction accuracy while the decision logic handles explanation, ensuring that neither function compromises the other. This segmentation prevents loss of interpretation information while maintaining high prediction accuracy.
Solution Approach 2:
The patent creates a simplified copy or representation of the prediction logic in the form of decision logic. This copy captures the essential reasoning pathways in a human-readable format, preserving interpretation information without requiring the full complexity of the original model.
3Reliability
If black-box prediction models are deployed, then predictive performance is improved, but user understanding and trust deteriorate
Solution Approach 1:
The decision logic serves as an intermediary that translates black-box predictions into understandable reasoning for users. This intermediary maintains the high predictive performance of the black-box model while improving user understanding by providing clear, interpretable explanations for each prediction.
Solution Approach 2:
The patent changes the representation parameters of the model output from opaque numerical predictions to structured decision logic with human-readable rules. This parameter change makes the model's reasoning process visible and understandable to users while preserving the underlying predictive performance.
Data Source
AI summary
A computer system is accessible to a database storing learning data to generate a prediction model, the learning data includes input data and teacher data, the computer system: performs first learning to set an extraction criterion for extracting the learning data including the input data similar to prediction target data in a case of being input the prediction target data; extract the learning data from the first database based on the extraction criterion and generate a dataset; perform second learning to generate a prediction model using the dataset; generate a decision logic showing a prediction logic of the prediction model; and output information to present the decision logic.