Table 4 Performance of the standalone L-SKRidge, dRVFL, LASSO, KRidge, and CFNN models based on assessment metrics in Brook River.

From: A hybrid framework: singular value decomposition and kernel ridge regression optimized using mathematical-based fine-tuning for enhancing river water level forecasting

 

Methods

Mode

R

RMSE

MAPE

NSE

IA

MaxAE

U95%

WL (t + 1)

L-SKRidge

Train

0.903

0.094

7.880

0.816

0.946

1.624

0.262

Test

0.893

0.092

8.145

0.797

0.941

0.915

0.254

dRVFL

Train

0.853

0.115

11.799

0.728

0.915

1.603

0.318

Test

0.853

0.110

11.030

0.706

0.897

1.070

0.303

LASSO

Train

0.844

0.118

11.616

0.712

0.909

1.707

0.327

Test

0.857

0.109

10.513

0.711

0.898

1.066

0.301

KRidge

Train

0.888

0.101

10.016

0.787

0.937

1.562

0.281

Test

0.873

0.099

9.875

0.762

0.926

1.051

0.275

CFNN

Train

0.893

0.099

8.905

0.797

0.939

1.676

0.275

Test

0.881

0.096

9.131

0.775

0.934

1.008

0.267

WL (t + 3)

L-SKRidge

Train

0.756

0.144

13.715

0.571

0.839

1.685

0.400

Test

0.726

0.141

12.778

0.519

0.818

1.177

0.389

dRVFL

Train

0.705

0.156

15.763

0.497

0.804

1.698

0.433

Test

0.705

0.146

14.376

0.482

0.788

1.157

0.403

LASSO

Train

0.702

0.157

15.716

0.492

0.800

1.682

0.435

Test

0.705

0.146

14.148

0.483

0.788

1.159

0.403

KRidge

Train

0.718

0.154

18.032

0.507

0.812

1.707

0.427

Test

0.712

0.143

15.275

0.506

0.804

1.161

0.396

CFNN

Train

0.726

0.152

16.633

0.524

0.828

1.558

0.420

Test

0.686

0.149

15.249

0.463

0.799

1.155

0.412

Back to article page