Disturbance Storm Time Index Prediction using Long Short-Term Memory Machine Learning

The cosmic matter that has the most influence on space weather on earth is greatly influenced b y s olar activity. Abnormal solar activity often affects the intensity of the solar wind into space, which is known as the geomagnetic storm phenomenon. One of the impacts caused by this phenomenon is the disruption of the satellite navigation system. In determining solar activity that affects the earth, observing the CME (Coronal Mass Eject) and flares c ontinuously i s n ecessary. O ne o f the references for measuring the level of geomagnetic storms is the disturbance storm time index (Dst-index). This paper predicts the Dst-index based on data from the OMNI web obtained from NASA's Advanced Composition Explorer (ACE) satellite. This paper aims to predict the disturbance storm time index using long short-term memory (LSTM). The results of the LSTM model were then evaluated using the root mean square error (RMSE) from the training results and testing results for comparative analyses of data with prediction to determine the error level. The best LSTM model for the Dst-index prediction shows the RMSEs are around the value of 3 for the training and testing.