Table 5 Results of performance evaluation of the six employed models in approaches I and II.

From: On the interpretability of machine and deep learning techniques for predicting CBR of stabilized soil containing agro-industrial wastes

Approach I

Approach II

MDL models

R

RMSE

MAE

RSD

U95

VAF

R

RMSE

MAE

RSD

U95

VAF

Training stage

Training stage

ANN

0.98

3.30

2.22

0.18

36.93

98.26

0.97

4.86

3.95

0.26

37.59

97.67

MARS

0.98

2.69

1.75

0.14

36.74

98.79

0.97

3.70

2.25

0.20

37.08

97.49

M5p-MT

0.98

3.33

1.86

0.18

36.94

97.78

0.97

4.03

2.36

0.22

37.21

96.88

LWP

0.94

6.02

3.04

0.32

38.23

92.12

0.97

3.82

2.51

0.21

37.13

97.58

XGBoost

0.99

2.14

0.99

0.12

36.60

98.95

0.98

2.67

1.42

0.14

36.74

98.51

LSTM

0.98

3.18

2.06

0.17

36.89

98.30

0.98

3.12

2.30

0.20

37.09

98.25

Testing stage

Testing stage

ANN

0.97

3.70

2.36

0.21

31.61

96.71

0.97

4.61

5.42

0.31

32.37

94.36

MARS

0.97

3.93

2.47

0.17

31.39

96.21

0.97

3.85

2.38

0.23

31.78

96.27

M5p-MT

0.97

3.54

1.90

0.21

31.62

96.38

0.97

4.31

2.96

0.26

31.94

96.03

LWP

0.96

4.94

2.51

0.38

33.12

92.67

0.96

4.82

3.07

0.24

31.84

94.11

XGBoost

0.98

3.17

1.70

0.24

31.23

96.84

0.97

3.88

2.25

0.27

31.38

95.27

LSTM

0.99

2.82

1.51

0.20

31.06

98.11

0.98

3.38

2.11

0.24

31.19

97.91

Back to article page