Ensemble deep learning models for prediction and uncertainty quantification of ground magnetic perturbation
Bibliographic record
Abstract
Geomagnetically Induced Currents are one of the most hazardous effects caused by geomagnetic storms. In the past literature, the variations in ground magnetic fields over time, dB/dt were used as a proxy value for GIC. Machine Learning (ML) techniques have emerged as a preferred methodology to predict dB/dt. However, space weather data are highly dynamic in nature, and the data distribution is subject to change over time due to environmental variability. The ML models developed are prone to the uncertainty in the input data and therefore suffer from high variance. In addition, a part of an ML architecture performance is conditional on the variables used to model the system in focus. Therefore, a single algorithm may not generate the required accuracy for a given dataset. In this work, a Bayesian Ensemble ML model has been developed to predict the variations over time of the local ground magnetic horizontal component, dBH/dt. The Ensemble methodology combines multiple ML models in the prediction process to predict dBH/dt. Bayesian statistics allow the estimation of model parameters and output as probability distributions, where the variance quantifies the uncertainty. The input data consists of solar-wind data from OmniWeb for the years 2001–2010. The local ground horizontal magnetic components for the corresponding time were calculated using SuperMAG data for the Ottawa ground magnetometer station for the years mentioned above. The years 2011–2015 were selected for model testing, as it encompasses the 5 August 2011 and 17 March 2015 geomagnetic storms. Five different accuracy metrics were considered; namely, Root Mean Squared Error (RMSE), Probability of Detection (POD), Probability of False Detection (PFD), Proportion Correct (PC), and Heidke Skills Score (HSS). The parameter uncertainty of the models is quantified, and the mean predicted dBH/dt is generated with a 95% credible interval. It can be observed that different models perform better with different datasets and the ensemble model has an accuracy comparable to the models with a relatively strong performance.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.002 | 0.005 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.002 | 0.001 |
| Research integrity | 0.001 | 0.002 |
| Insufficient payload (model declined to judge) | 0.001 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".