Multiple‐Hour‐Ahead Forecast of the Dst Index Using a Combination of Long Short‐Term Memory Neural Network and Gaussian Process

Abstract

In this study, we present a method that combines a Long Short-Term Memory (LSTM) recurrent neural network with a Gaussian process (GP) model to provide up to 6-hr-ahead probabilistic forecasts of the Dst geomagnetic index. The proposed approach brings together the sequence modeling capabilities of a recurrent neural network with the error bars and confidence bounds provided by a GP. Our model is trained using the hourly OMNI and Global Positioning System (GPS) databases, both of which are publicly available. We first develop a LSTM network to get a single-point prediction of Dst. This model yields great accuracy in forecasting the Dst index from 1 to 6 hr ahead, with a correlation coefficient always higher than 0.873 and a root-mean-square error lower than 9.86. However, even if global metrics show excellent performance, it remains poor in predicting intense storms (Dst < À250 nT) 6 hr in advance. To improve it and to obtain probabilistic forecasts, we combine the LSTM model obtained with a GP and evaluate the hybrid predictor using the receiver operating characteristic curve and the reliability diagram. We conclude that this hybrid methodology provides improvements in the forecast of geomagnetic storms, from 1 to 6 hr ahead.

Dst Index Forecast Based on Ground‐Level Data Aided by Bio‐Inspired Algorithms

“…A direct comparison between the results obtained by the proposed methods and other methods reported in the literature (as presented in Table ) is complicated due to differences in databases and methodologies used in each study (Andriyas & Andriyas, ; Andrejková & Levický, ; Bala & Reiff, ; Barkhatov et al, ; Gleisner et al, ; Gruet et al, ; Jankovičová et al, ; Kugblenu et al, ; Lazzús et al, ; Lethy et al, ; Lotfi & Akbarzadeh‐T., ; Lundstedt & Wintoft, ; Lundstedt et al, ; Munsami, ; Ouarbya & Mirikitani, ; Ouarbya et al, ; Pallocchia et al, ; Revallo et al, ; Sharifi et al, ; Sharifi et al, ; Singh & Singh, ; Stepanova et al, ; Stepanova & Pérez, ; Stepanova et al, ; Vega‐Jorquera et al, ; Vörös & Jankovičová, ; Watanabe et al, ; ; Wei et al, ; Wu & Lundstedt, ; ; Xue & Gong, ). In addition, the ANN configurations contain deep variation such as the time steps of ahead prediction (i.e., output), input parameters, and the number of neurons in the hidden layer (see Table ).…”

The Role of Global/Regional Earth Conductivity Models in Natural Geomagnetic Hazard Mitigation

“…7). Another promising avenue for prediction of coronal mass ejections on the Sun and related geomagnetic field disturbances involves machine learning techniques (e.g., Bobra and Ilonidis 2016;Gruet et al 2018;Camporeale et al 2018), providing an alternative approach for future GIC forecasts.…”

Reconstruction of Time Series Observed in Linear Magnetized Plasma PANTA via a Machine Learning Algorithm

“…Recently, recurrent neural networks [11][12][13], in which past information of the time series is also used for prediction, have been applied to chaotic dynamical systems [11], geomagnetic activity [12] and nuclear fusion [13]. The machine learning algorithm has also been applied to improve turbulence modeling [14,15].…”