Skip to main content

Table 3 Stage 1, 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

From: A comparison of machine learning methods for ozone pollution prediction

Model

rMSE \(\downarrow\)

MAE \(\downarrow\)

MAPE \(\downarrow\)

R-square \(\uparrow\)

J2 \(\downarrow\)

Time (min)

Linr

12.677/16.196

10.09/14.137

0.945/2.713

0.332/0.305

1.632

0.01

Linr_l2

12.677/16.199

10.09/14.141

0.945/2.714

0.332/0.305

1.633

0

Lasso

13.642/19.561

10.91/17.674

1.114/3.589

0.227/0.267

2.056

0

PLSR

12.777/17.336

10.243/15.319

0.934/2.901

0.322/0.331

1.841

0.01

GRP_Expo

9.117/12.612

6.847/10.237

0.543/1.941

0.655/0.387

1.914

9.64

GRP_DotProd

12.706/16.754

10.121/14.705

0.963/2.847

0.329/0.306

1.739

121.44

GRP_Matern

0.0/20.868

0.0/17.887

0.0/1.0

1.0/0.0

inf

25.47

SVR_linear

12.795/15.982

9.99/13.821

0.997/2.69

0.323/0.218

1.56

0.01

SVR_poly

11.858/13.271

9.358/10.827

0.873/2.002

0.416/0.177

1.253

5.45

SVR_rbf

10.66/13.062

8.202/10.981

0.685/2.007

0.528/0.428

1.501

8.57

SVR_sigmoid

12.764/16.48

10.124/14.4

0.973/2.778

0.323/0.278

1.667

9.77

MLP_1

12.681/16.561

10.111/14.523

0.939/2.755

0.332/0.316

1.706

0.6

MLP_2

15.514/21.591

12.682/19.384

1.283/4.199

0.0/0.0

1.937

1.35

RF

10.116/13.874

7.801/11.75

0.657/2.188

0.575/0.42

1.881

0.06

Bagging

3.722/13.254

2.497/10.781

0.174/2.121

0.942/0.363

12.681

0.18

GBoost

10.272/14.359

7.948/12.269

0.66/2.306

0.562/0.398

1.954

0.28

AdaBoost

11.704/17.821

9.465/16.113

0.852/3.047

0.431/0.449

2.318

0.1

HistGBoost

9.036/13.009

6.899/10.831

0.538/2.147

0.661/0.431

2.073

13.07

LightGBM

11.532/16.499

9.204/14.699

0.892/3.051

0.447/0.422

2.047

0.01

  1. In addition, the \(\uparrow\) means the model is better when the value is larger, and the \(\downarrow\) means the model is better when the value is smaller