Design of Deep Belief Networks for Short-Term Prediction of Drought Index Using Data in the Huaihe River Basin

Table 4

The comparison of RMSE and MAE between BP and DBN.

Station	Model	Errors	SPI3	SPI6	SPI9	SPI12
Bengbu	DBN	RMSE	0.6842	0.6592	0.5355	0.4797
		MAE	0.5425	0.5274	0.3959	0.3553
	BP neural network	RMSE	0.9897	0.6987	0.5899	0.5809
		MAE	0.7564	0.5523	0.4157	0.4532
Fuyang	DBN	RMSE	0.8112	0.6634	0.5590	0.5620
		MAE	0.6527	0.4812	0.3923	0.4282
	BP neural network	RMSE	1.0876	0.8022	0.8032	0.5773
		MAE	0.8202	0.5867	0.5509	0.4080
Xuchang	DBN	RMSE	0.7258	0.5764	0.5262	0.4236
		MAE	0.5714	0.4342	0.3880	0.2976
	BP neural network	RMSE	0.8223	0.6938	0.6783	0.4454
		MAE	0.6786	0.5411	0.4725	0.3268
Zhumadian	DBN	RMSE	0.7794	0.6239	0.5686	0.4990
		MAE	0.6276	0.4792	0.3811	0.3504
	BP neural network	RMSE	1.0780	0.7956	0.7996	0.4474
		MAE	0.8147	0.6336	0.5365	0.3144