COVID-19 Pandemic Forecasting Using CNN-LSTM: A Hybrid Approach

Table 4

The average model performance evaluation for forecasting confirmed cases on forecast data.

Type	Models	MAPE	RMSE	RRMSE
Proposed approach	CNN-LSTM	0.43	8780.71	3.01
Deep learning	CNN-1D	0.59	12349.46	4.16
	LSTM	0.86	17257.32	6.01
Statistical method	ARIMA	0.82	16156.94	5.73
	FBProphet	0.83	17223.50	5.82
Linear model	LR	1.70	34688.47	11.88
	Ridge	1.31	27657.47	9.20
	Lasso	1.70	34669.07	11.88
Ensemble	XGBR	1.40	28970.67	9.83
	AdaBoostR	1.66	33822.49	11.63
	RFR	1.64	33395.34	11.51
	GBR	1.73	34770.86	12.12
	ETR	1.50	30593.47	10.47
	BaggingR	1.64	33337.85	11.48
Machine learning	GPR	1.65	33878.38	11.58
	SVR	1.35	27944.14	9.48
	DTR	1.75	35680.97	12.24
	KNNR	1.55	31464.57	10.82

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 forecast data.