Table 6 Outstanding performance by the WIO-SVR model.

From: Proposing a hybrid metaheuristic optimization algorithm and machine learning model for energy use forecast in non-residential buildings

Statistical indices

Intelligent prediction model

Accuracy change by WIO-SVR model

WIO-SVR

SVR

RF

M5P

REPTree

SVR

RF

M5P

REPTree

RMSE (kWh)

Average

2.02

10.95

16.27

17.73

26.44

442.0a

705.0a

777.4a

1207.9a

Std. dev

1.10

10.24

15.59

15.84

20.43

    

MAE (kWh)

Average

5.25

7.55

11.33

12.29

17.77

43.7c

115.6a

134.0a

238.2a

Std. dev

4.61

6.82

10.30

10.27

15.02

    

MAPE (%)

Average

15.5

26.2

46.4

45.3

78.7

69.1a

199.3a

192.4a

407.8a

Std. dev

8.0

15.1

38.0

28.9

79.0

    

R

Average

0.90

0.86

0.84

0.74

0.69

4.3b

6.0a

17.4a

22.9a

Std. dev

0.05

0.12

0.16

0.16

0.24

    
  1. aIndicates the significant level less than 1%.
  2. bIndicates the significant level less than 5%.
  3. cIndicates the significant level less than 10%.