Table 4.

Predictive performance of the combinations of feature sets.

LeadTime (day)/ feature sets	Metrics	LIWC	Topic	Temporal features of the OPHS	COVID-19 cases	The combination of all four feature sets	Optimal features combination	Performance of the optimal combination	Prophet
1	MAE	7.758	7.392	6.550	11.882	6.265	Topic & Timeseries & Covid19 pandemic	6.211	5.93
	Pearson Coef	0.731**	0.781**	0.798**	0.822**	0.884**		0.898**	0.923**
3	MAE	8.149	7.705	6.214	13.396	5.929	Topic & timeseries	5.780	5.96
	Pearson Coef	0.876**	0.898**	0.885**	0.906**	0.911**		0.932**	0.911**
7	MAE	8.258	8.343	6.470	11.947	6.400	Topic & timeseries	6.223	6.34
	Pearson Coef	0.916**	0.876**	0.913**	0.913**	0.924**		0.940**	0.908**
14	MAE	9.258	8.881	8.416	12.953	7.779	LIWC & topic & Timeseries & Covid-19 pandemic	7.779	5.92
	Pearson Coef	0.928**	0.93**	0.931**	0.901**	0.942**		0.942**	0.903**

^**p < 0.001.