Research on the Combined Prediction Model of Residential Building Energy Consumption Based on Random Forest and BP Neural Network

Table 5

Comparison of RF-BPNN with existing model results.

	Methods	MAE	RMSE	MAPE (%)
HL	IRLS [15]	2.1400	3.1400	10.0900
	SVR [16]	0.4320	0.6100	—
	CART [16]	0.4370	0.8000	—
	ANN-SVR [7]	0.3000	0.4280	1.5570
	HYBRID-LIN [6]	0.5100	0.7874	0.4700
	KNN	1.9529	2.3329	8.4504
	RF-BPNN	0.3199	0.4550	1.4591
CL	IRLS [15]	2.2100	3.3900	8.4100
	SVR [16]	0.8900	1.6470	—
	CART [16]	1.1570	1.8410	—
	ANN-SVR [7]	0.973	1.5660	3.4550
	HYBRID-LIN [6]	1.1800	2.0372	3.3300
	KNN	1.8193	2.2651	7.1413
	RF-BPNN	0.7765	1.1614	3.0000