Explainable prediction of daily hospitalizations for cerebrovascular disease using stacked ensemble learning

Table 3 Performance comparison of different methods in predicting HAs for CD and stroke

Datasets	Models	MAE	RMSE	MAPE	R²
CD	RF	14.713	20.649	0.154	0.652
	GBDT	14.661	20.296	0.154	0.663
	Ridge	14.894	19.963	0.183	0.674
	ANN	14.408	18.407	0.191	0.723
	LSTM	13.774	18.421	0.165	0.739
	Stacking	12.467	17.053	0.153	0.762
	Stacking + LDS	11.855^*	16.078^*	0.145	0.789
Stroke	RF	10.889	14.660	0.175	0.51
	GBDT	11.251	14.995	0.178	0.487
	Ridge	10.357	13.278	0.210	0.598
	ANN	9.525	12.140	0.191	0.664
	LSTM	9.422	12.166	0.185	0.676
	Stacking	9.038	11.898	0.170	0.677
	Stacking + LDS	8.961^*	11.850	0.159^*	0.680

The best result for each metric is in bold.
^*The differences in the MAE, RMSE, or MAPE between the stacking model with LDS and the best individual model are significant (P-value < 0.05) according to the t-test

ISSN: 1472-6947