Magnetic substorms which are caused by the
interaction of the earth magnetospheric and ionospheric layers
with the solar wind may negatively impact the satellite
communication systems. Several research studies have used the
SuperMAG Lower Index (i.e., SML-Index) to detect the onset of
the magnetic substorms. It is also worth mentioning that the use of
machine and deep learning for the prediction of the SML-Index
has not yet been explored in the literature. In this paper, we
proposed a novel deep learning framework named the SML-Net
that was able to predict future ten-minute time-frames of the SMLIndex
based on the trend, seasonality and residual components of
past sixty-minute time-frames of the same index at a relatively low
normalized root mean squared error of 0.34 and comparatively
high coefficient of determination of 92.9% with respect to four
benchmark models including the stacked long short term memory
(LSTM), the stacked gated recurrent unit (GRU), the temporal
convolutional network (TCN) and the transformer encoder