Table 1. Performance of dengue incidence prediction models from different time horizons, for the time period between January 2011 and July 2016, in the city of Barra Mansa, State of Rio de Janeiro, Brazil.

Each time horizon is examined across all four possible features sets: autoregressive terms alone (AR), autoregressive terms together with Google Trends data (AR+GT) and with weather data (AR+GT+W), as well as google trends data alone (GT). Numbers in bold represent the best performance for a given model and autoregressive lag across each of the metrics. This corresponds to the lowest value for the RMSE and relative RMSE metrics, and the highest value for the R^2 and Pearson correlation metrics.

Model	Reporting Delay	Features	RMSE	Relative RMSE	R^2	Pearson Correlation
Lasso Regression	8 weeks	AR	28.425	0.897	0.009	0.188
		GT	23.606	0.745	0.317	0.563
		AR+GT	23.828	0.752	0.304	0.555
		AR+GT+W	24.67	0.778	0.254	0.505
	6 weeks	AR	26.65	0.845	0.122	0.355
		GT	23.546	0.746	0.315	0.562
		AR+GT	22.615	0.717	0.368	0.608
		AR+GT+W	23.039	0.73	0.344	0.587
	3 weeks	AR	19.264	0.615	0.536	0.733
		GT	21.581	0.689	0.418	0.649
		AR+GT	18.859	0.602	0.556	0.752
		AR+GT+W	18.789	0.6	0.559	0.753
	1 week	AR	12.485	0.4	0.804	0.897
		GT	20.229	0.649	0.485	0.703
		AR+GT	12.259	0.393	0.811	0.901
		AR+GT+W	12.222	0.392	0.812	0.901
Random Forest	8 weeks	AR	26.382	0.832	0.146	0.508
		GT	21.061	0.664	0.456	0.68
		AR+GT	23.355	0.737	0.331	0.596
		AR+GT+W	22.514	0.71	0.378	0.631
	6 weeks	AR	24.625	0.781	0.251	0.591
		GT	22.203	0.704	0.391	0.642
		AR+GT	22.055	0.699	0.399	0.664
		AR+GT+W	21.154	0.671	0.447	0.678
	3 weeks	AR	18.332	0.585	0.58	0.776
		GT	21.07	0.673	0.445	0.676
		AR+GT	17.613	0.562	0.612	0.793
		AR+GT+W	19.354	0.618	0.532	0.749
	1 week	AR	11.047	0.354	0.846	0.92
		GT	19.408	0.622	0.526	0.729
		AR+GT	11.027	0.354	0.847	0.924
		AR+GT+W	11.844	0.38	0.823	0.91