Wavelet Predictor Variable Data for Explainable Machine Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current machine learning technologies face challenges in accurately predicting the timing of adverse events from time series data, particularly in handling missing values and providing clear explanations for predictions, which limits their effectiveness in modifying machine-implemented operating environments.
Innovation Solution
The use of wavelet transforms to generate predictor variable data, combined with multiple modeling algorithms and techniques like points below value, integrated gradients, and Shapley values, to improve prediction accuracy and explainability, allowing for more effective timing predictions and adaptive operating environment modifications.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If wavelet transforms are applied to generate predictor variable data, then prediction accuracy is improved, but computational complexity increases
Solution Approach 1:
The wavelet transform is applied in advance to the time series data to generate wavelet predictor variables before the main prediction process. This preprocessing step decomposes the original data into multiple scales and shifts, creating enriched feature representations that improve prediction accuracy while allowing the main model to focus on learning patterns from these pre-processed features
2Measurement precision
If multiple modeling algorithms are used to compute probabilities for different time windows, then prediction accuracy is improved, but model complexity increases
Solution Approach 1:
The prediction task is segmented into multiple time windows, with a separate modeling algorithm trained for each time window. Each algorithm computes probabilities for its specific time window, allowing the system to capture temporal patterns at different scales. This segmentation improves accuracy by addressing the temporal specificity of events while organizing complexity into manageable, independent components
Solution Approach 2:
The set of modeling algorithms serves multiple functions: each algorithm predicts probabilities for a specific time window, and collectively they provide comprehensive timing predictions across the entire time horizon. This multi-functionality allows a single system to handle diverse prediction requirements without requiring separate specialized systems for each time window
3Loss of information
If explainable machine-learning techniques are implemented, then interpretability is improved, but computational overhead increases
Solution Approach 1:
Explainable machine-learning techniques serve as intermediaries between the complex multi-algorithm prediction system and the end user. These techniques translate the outputs from multiple modeling algorithms into interpretable explanations that reveal which features and time windows contribute most to predictions, reducing information loss about model decision-making while managing computational overhead through efficient explanation generation methods
Data Source
AI summary
A host computing system determines a wavelet transform that represents time-series values of predictor data samples. The host computing system applies the wavelet transform to the predictor data samples to generate wavelet predictor variable data comprising a first set and a second set of shift value input data for a first scale and a second scale. The host computing system computes a set of probabilities for a target event by applying a set of timing-prediction models to the first set and the second set of shift value input data. The host computing system determines an event prediction from the set of probabilities and modifies a host system operation based on the determined event prediction.