COVID-19 Pandemic Forecasting Using CNN-LSTM: A Hybrid Approach
Table 3
The average model performance evaluation for forecasting confirmed cases on test data.
Type
Models
MAPE
RMSE
RRMSE
Proposed approach
CNN-LSTM
0.19
13275.00
5.30
Deep learning
CNN-1D
0.23
15954.97
6.34
LSTM
0.26
17512.45
7.20
Statistical method
ARIMA
0.20
13630.13
5.51
FBProphet
0.33
22326.90
9.25
Linear model
LR
0.35
24150.13
9.82
Ridge
0.34
24786.26
9.65
Lasso
0.35
24143.65
9.82
Ensemble
XGBR
0.30
21267.73
8.51
AdaBoostR
0.30
21065.30
8.39
RFR
0.31
21433.81
8.56
GBR
0.30
21140.69
8.49
ETR
0.28
19210.06
7.73
BaggingR
0.31
21435.07
8.55
Machine learning
GPR
0.33
22461.64
9.18
SVR
0.76
56131.22
21.18
DTR
0.33
22954.19
9.15
KNNR
0.26
18954.48
7.28
The bold values present the lowest error values of MAPE, RMSE, and RRMSE. These values show that the proposed approach outperforms the baseline models based on the test data.