Prediction and analysis of COVID-19 daily new cases and cumulative cases: times series forecasting and machine learning models

Table 3 Accuracy evaluation of ARIMA, SARIMA and Prophet on fitting and forecasting COVID-19 in USA, Brazil and India

	County	Model	Model fitting				Model forecasting
	County	Model	R²	RMSE	MAE	MAPE	R²	RMSE	MAE	MAPE
New Cases	USA	SARIMA	0.942	14,850.734	7877.085	58.421	0.376	100,008.409	67,381.862	10.022
	USA	Prophet	0.950	13,437.603	7118.961	59.665	0.485	67,842.843	50,774.029	10.672
	Brazil	SARIMA	0.821	10,145.700	5661.420	101.147	0.277	2998.022	2490.761	18.477
	Brazil	Prophet	0.850	9305.905	4819.421	84.732	0.185	5432.666	4593.246	36.437
	India	SARIMA	0.997	5807.807	2847.320	30.662	0.254	1648.759	975.779	3.371
	India	Prophet	0.997	4073.903	2331.656	29.744	0.244	2697.648	2324.177	8.747
Cumulative Cases	USA	ARIMA	1.000	17,702.819	10,404.341	0.431	0.972	1,149,640.000	903,809.070	0.530
	USA	Prophet	0.745	10,616,818.000	7,862,825.100	242.943	0.084	982,545.500	623,199.520	0.362
	Brazil	ARIMA	1.000	13,046.465	8325.811	1.099	0.949	53,256.662	38,376.992	0.052
	Brazil	Prophet	1.000	46,232.061	30,472.587	2.072	0.885	985,393.238	784,833.600	1.060
	India	ARIMA	1.000	7057.269	4189.751	0.439	0.999	10,525.591	9883.822	0.009
	India	Prophet	1.000	43,676.607	26,233.945	1.430	0.880	390,438.575	263,797.980	0.228

The R²-value of 1.0 on the graph means that the correlation coefficient is greater than 0.9995, approximately 1.0

ISSN: 1471-2334