Table 4 Stage 2, the comparison of machine learning methods on ozone prediction, where the value before slash represents the loss value in the training dataset and the latter in the testing dataset (training/testing), the best performance is bold and the line colored grey means over-fitting

Linr

9.847/11.958

7.634/10.116

0.484/1.052

0.597/-0.111

1.475

0.01

Linr_l2

9.847/11.956

7.635/10.114

0.484/1.051

0.597/-0.11

1.474

0

Lasso

11.035/13.558

8.867/12.345

0.77/2.111

0.494/0.379

1.51

0.01

PLSR

9.956/12.148

7.761/10.339

0.496/1.067

0.588/0.013

1.489

0.01

GRP_Expo

7.558/9.968

5.696/7.383

0.31/1.051

0.763/0.302

1.739

9.04

GRP_DotProd

9.848/11.94

7.635/10.09

0.482/1.049

0.597/-0.111

1.47

132.69

GRP_Matern

0.0/20.868

0.0/17.887

0.0/1.0

1.0/0.0

inf

25.77

SVR_linear

9.901/12.106

7.582/9.98

0.464/1.075

0.593/-0.247

1.495

0.01

SVR_poly

9.937/11.656

7.663/9.096

0.56/0.949

0.59/-0.141

1.376

6.82

SVR_rbf

8.54/9.19

6.437/7.302

0.38/1.047

0.697/0.499

1.158

9.24

SVR_sigmoid

9.947/12.026

7.69/10.106

0.484/1.052

0.589/-0.163

1.462

10.24

MLP_1

8.821/9.505

6.703/7.642

0.41/1.056

0.677/0.416

1.161

1.33

MLP_2

8.809/9.985

6.704/8.022

0.397/1.23

0.678/0.457

1.285

1.42

RF

8.059/9.921

6.153/7.874

0.357/0.953

0.73/0.436

1.515

0.06

Bagging

2.919/10.164

1.94/7.99

0.097/0.918

0.965/0.377

12.124

0.1

GBoost

8.191/9.856

6.218/7.926

0.366/0.863

0.721/0.394

1.448

0.26

AdaBoost

9.413/13.832

7.46/12.113

0.561/1.758

0.638/0.445

2.159

0.11

HistGBoost

7.214/9.509

5.458/7.53

0.305/0.84

0.784/0.419

1.737

8.42

LightGBM

9.712/12.974

7.721/11.671

0.69/2.167

0.608/0.586

1.785

0.01