White Box Models Explain Black Box Time Series Predictions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Black box time series models lack explainability, making it difficult to understand and debug their predictions, which can lead to errors and a lack of trust in the model.
Innovation Solution
The method involves using training data and similar white box time series models to explain predictions made by black box time series models. This is achieved by generating a set of white box models similar to the black box model, analyzing their behavior, and creating an explanation based on the training data and preferred white box model.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If black box time series models are used for predictions, then prediction accuracy may be improved, but explainability of predictions deteriorates
Solution Approach 1:
The patent introduces white box time series models as intermediary elements that mediate between the black box model's predictions and the need for explainability. These white box models serve as transparent proxies that replicate the black box model's behavior while providing interpretable reasoning for predictions, thus resolving the contradiction between accuracy and explainability.
Solution Approach 2:
The patent creates copies of the black box model's predictive behavior using white box models. By generating multiple white box models that mimic the black box model's predictions and selecting the best matching one, the system preserves the predictive accuracy while enabling explanation through the transparent structure of white box models.
2Reliability
If black box time series models are used, then predictive power is improved, but debugging capability deteriorates
Solution Approach 1:
White box models serve as intermediary representations that allow debugging of black box model predictions. By creating transparent proxies that replicate black box behavior, developers can inspect the reasoning process, identify sources of incorrect predictions, and debug the model without losing predictive power.
Solution Approach 2:
Instead of trying to make the black box model itself transparent, the patent inverts the approach by creating white box models that replicate black box predictions. This inversion allows debugging through the transparent white box models while preserving the black box model's predictive capabilities.
3Productivity
If black box models are deployed, then forecasting performance is improved, but trust in model predictions deteriorates
Solution Approach 1:
The white box models act as intermediaries that provide trustworthy explanations for black box predictions. By offering transparent reasoning that can be verified and understood, these intermediaries build trust in the forecasting performance delivered by the black box model without compromising its accuracy.
Solution Approach 2:
The patent implements a feedback mechanism where white box model explanations provide insight into black box predictions. This feedback loop allows users to understand and verify predictions, thereby building trust while maintaining high forecasting performance.
Data Source
AI summary
A method, system, and computer program product for explaining predictions made by black box time series models. The method may include identifying a black box time series model. The method may also include predicting one or more time instances using the black box time series model. The method may also include selecting a predicted time instance from the predicted data. The method may also include receiving training data for the black box time series model. The method may also include generating a set of white box time series models similar to the black box time series model. The method may also include selecting a preferred white box time series model. The method may also include analyzing behavior of the preferred white box time series model. The method may also include generating an explanation illustrating why the black box time series model forecasted the predicted time instance.