Table 5 Results of a performance evaluation using the developed ML models during training, testing and validation phase for maximum air temperature forecasting.

Model	Date set	NSE	d	MAE	RMSE	RAE	RRSE	PCC	R²
Linear Regression	Training	0.8708	0.9644	1.9981	2.7828	29.7166	35.945	0.9332	0.8708
	Testing	0.8466	0.9574	2.0832	2.9090	32.0717	38.815	0.9202	0.8467
	Validation	0.8705	0.9643	2.0001	2.7861	29.7428	35.985	0.9330	0.8705
Additive Regression	Training	0.8362	0.9529	2.3187	3.1329	34.4855	40.467	0.9147	0.8367
	Testing	0.8072	0.9445	2.4270	3.2611	37.3644	43.512	0.8989	0.8079
	Validation	0.8305	0.9520	2.3565	3.1872	35.0419	41.165	0.9113	0.8305
Random Subspace	Training	0.8675	0.9618	2.0719	2.8181	30.8150	36.402	0.9323	0.8693
	Testing	0.8231	0.9481	2.2817	3.1238	35.1278	41.680	0.9080	0.8244
	Validation	0.8209	0.9570	2.1836	2.9890	32.4712	38.606	0.9230	0.8519
M5P	Training	0.8761	0.9660	1.9555	2.7248	29.0838	35.196	0.9360	0.8761
	Testing	0.8473	0.9578	2.0773	2.9027	31.9806	38.731	0.9206	0.8475
	Validation	0.8720	0.9649	1.9867	2.7696	29.5440	35.773	0.9338	0.8720
SVM	Training	0.8696	0.9645	1.9861	2.7952	29.5388	36.106	0.9328	0.8701
	Testing	0.8453	0.9575	2.0756	2.9215	31.9543	38.982	0.9199	0.8463
	Validation	0.8694	0.9644	1.9881	2.7975	29.5648	36.133	0.9327	0.8699

The best-performing machine learning model for forecasting daily minimum air temperature is highlighted in bold blue, while the poorest-performing model is indicated in red.

Quick links

Search