Predicting Multidimensional Environmental Factor Trends in Greenhouse Microclimates Using a Hybrid Ensemble Approach

Table 3

Prediction errors of combined CEEMDAN models.

Method	Metric	CO2 concentration	Atmospheric pressure	Light intensity	Humidity	Temperature
CEEMDAN-RNN	MSE	526.682	708.483	317257.496	1.506	0.618
	RMSE	22.950	26.617	563.256	1.227	0.786
		0.975	0.996	0.961	0.991	0.974
	MAE	12.987	19.175	283.962	0.853	0.597
	MAPE	2.077%	0.020%	——	1.718%	7.483%
	SMAPE	2.094%	0.020%	——	1.685%	7.249%
CEEMDAN-CNN	MSE	501.896	674.359	293723.387	1.479	0.473
	RMSE	22.403	25.968	541.963	1.216	0.688
		0.976	0.996	0.964	0.991	0.978
	MAE	12.731	18.046	271.895	0.807	0.491
	MAPE	1.896%	0.019%	——	1.687%	7.175%
	SMAPE	1.893%	0.019%	——	1.680%	6.874%
CEEMDAN-LSTM	MSE	514.781	667.048	246108.639	1.184	0.422
	RMSE	22.689	25.827	496.093	1.088	0.649
		0.977	0.997	0.969	0.992	0.981
	MAE	12.890	17.337	247.983	0.743	0.468
	MAPE	1.976%	0.017%	——	1.581%	6.755%
	SMAPE	1.967%	0.017%	——	1.593%	6.431%
CEEMDAN-BiLSTM	MSE	491.188	429.587	278586.624	1.038	0.380
	RMSE	22.163	20.726	527.813	1.019	0.616
		0.979	0.997	0.970	0.994	0.986
	MAE	12.531	14.986	258.615	0.687	0.423
	MAPE	1.890%	0.015%	——	1.378%	5.132%
	SMAPE	1.918%	0.015%	——	1.364%	4.897%
CNN-LSTM	MSE	334.845	294.584	182547.552	0.867	0.314
	RMSE	18.299	17.163	427.256	0.931	0.560
		0.986	0.997	0.978	0.995	0.988
	MAE	11.816	12.783	227.253	0.646	0.391
	MAPE	1.791%	0.013%	——	1.252%	4.876%
	SMAPE	1.792%	0.013%	——	1.258%	4.624%
CEEMDAN-Informer	MSE	358.043	242.237	100222.767	0.697	0.249
	RMSE	18.922	15.564	316.580	0.835	0.499
		0.984	0.998	0.986	0.996	0.991
	MAE	11.874	11.098	203.045	0.617	0.346
	MAPE	1.795%	0.011%	——	1.180%	4.250%
	SMAPE	1.795%	0.011%	——	1.185%	4.108%