Table 5 Model performance indicators

From: Full-cycle prediction of crack healing in self-healing concrete using generalized polynomial chaos expansion

Type

Age

q-norm

LOOCV

Mean value

Standard deviation

Coef. of variation

First-Order PCE

28

1.00

0.5395

77.8423

1.4890

1.279%

14

1.00

0.0646

70.4410

1.0488

1.489%

7

1.00

1.5653

50.0147

0.3198

0.639%

Second-Order PCE

28

1.00

3.6822

77.5307

1.5260

1.968%

14

1.00

0.0952

72.7136

2.8331

3.896%

7

1.00

2.2316

49.9524

1.2490

2.500%

SHG

28

1.00

0.7081

66.8720

7.6376

11.421%

  1. LOOCV (Leave-One-Out Cross-Validation) is a model validation metric calculated by iteratively training on all-but-one sample and predicting the excluded one, with lower values indicating better generalizability. q-norm represents the order of the norm used in the polynomial basis orthogonalization, where q = 1.00 indicates the L1-norm was employed for enhanced robustness to outliers.
Back to article page