In this article, a new hybrid time series model is proposed to predict COVID-19 daily
confirmed cases and deaths. Due to the variations and complexity in the data, it is
very difficult to predict its future trajectory using linear time series or mathematical
models. In this research article, a novel hybrid ensemble empirical mode decomposition
and error trend seasonal (EEMD-ETS) model has been developed to forecast the
COVID-19 pandemic. The proposed hybrid model decomposes the complex, nonlinear,
and nonstationary data into different intrinsic mode functions (IMFs) from low to high
frequencies, and a single monotone residue by applying EEMD. The stationarity of each
IMF component is checked with the help of the augmented Dicky–Fuller (ADF) test and
is then used to build up the EEMD-ETS model, and finally, future predictions have been
obtained from the proposed hybrid model. For illustration purposes and to check the
performance of the proposed model, four datasets of daily confirmed cases and deaths
from COVID-19 in Italy, Germany, the United Kingdom (UK), and France have been
used. Similarly, four different statistical metrics, i.e., root mean square error (RMSE),
symmetric mean absolute parentage error (sMAPE), mean absolute error (MAE), and
mean absolute percentage error (MAPE) have been used for a comparison of different
time series models. It is evident from the results that the proposed hybrid EEMD-ETS
model outperforms the other time series and machine learning models. Hence, it is
worthy to be used as an effective model for the prediction of COVID-19. |
