Hellinger Distance for DyBM Prediction Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for measuring the accuracy of Dynamic Boltzmann Machine (DyBM) predictions for both mean and standard deviation in financial time-series datasets, such as Negative Log-Likelihood (NLL), fail to account for the correlation between actual and predicted standard deviation, leading to potential losses in market opportunities.
Innovation Solution
Employing the Hellinger Distance (HD) to assess the accuracy of DyBM predictions for both mean and standard deviation, which takes into account the period during which the DyBM can accurately predict both values, and using this metric to determine a trustworthy prediction time period for informed decision-making.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If Negative Log-Likelihood (NLL) is used to measure prediction accuracy, then the method is simple and standard, but it fails to account for the correlation between actual and predicted standard deviation, leading to inaccurate assessment
Solution Approach 1:
The patent changes the measurement parameter from NLL to Hellinger Distance. The Hellinger Distance explicitly incorporates the correlation between actual and predicted standard deviation through its mathematical formulation, which compares probability distributions. This parameter change resolves the contradiction by providing a more precise measurement that captures the relationship between predicted and actual volatility, while remaining computationally feasible.
2Reliability
If NLL is used to penalize predictions during high volatility periods, then the penalty is applied uniformly, but this leads to loss of opportunities to obtain better yields by ignoring the correctness of predictions
Solution Approach 1:
The patent applies local quality by making the penalty mechanism differentiated based on the specific characteristics of each prediction period. Instead of uniform penalty, the Hellinger Distance evaluates the degree of mismatch between predicted and actual distributions, allowing for nuanced assessment that rewards accurate predictions even during high volatility periods while penalizing only when the distributional mismatch is significant.
3Adaptability or versatility
If a DyBM predicts both mean and standard deviation, then the model provides comprehensive information, but it becomes difficult to measure the accuracy of the combined prediction tuple
Solution Approach 1:
The patent merges the measurement of mean and standard deviation accuracy into a single unified metric - the Hellinger Distance. This distance metric simultaneously evaluates both components of the prediction tuple by comparing the full probability distributions, thereby simplifying the measurement process while maintaining comprehensive assessment of both mean and volatility predictions.
Data Source
AI summary
A method is provided for commodity management. The method generates, using a Dynamic Boltzmann Machine (DyBM), a future mean prediction and a future standard deviation prediction of a financial time-series dataset for a commodity. The method measures, using Hellinger Distance (HD), an accuracy of the future mean prediction and the future standard deviation prediction. The method combines the future mean prediction and the future standard deviation prediction with the Hellinger Distance to determine a DyBM trustworthy prediction time period in which predictions by the DyBM, including the future mean prediction and the future standard deviation prediction, are deemed trustworthy. The method selectively performs an action relating to an ownership of the commodity based on at least one of the future mean prediction and the future standard deviation prediction, responsive to the future mean prediction and the future standard deviation prediction being generated during the DyBM trustworthy prediction time period.