Table 3 Hyper parameter screening for conventional models

Hyper parameters

C

Kernel function

\(\gamma\)

Degree

Screened range

[0.5, 1.0]

RBF, Sigmoid, Polynomial

Scale, auto

[2, 6]

Best

1.0

RBF

Scale

N/A

Hyper parameters

# estimators

Max depth

Max features

Min samples per leaf

Screened range

[100, 2000]

[5, 60]

[11, 2048]

[1, 20]

Best

2000

20

200

1

Hyper parameters

# estimators

Max depth

Max features

Learning rate

Screened range

[500, 2000]

[15, 35]

[512, 2048]

[0.1, 0.5]

Best

500

25

1024

0.5

Hyper parameters

\(\lambda\)

\(\gamma\)

Max depth

# rounds

Learning rate (\(\varepsilon\))

Screened range

[0, 3]

[0, 16]

[5, 25]

[10, 25]

[0.1, 0.5]

Best

1

1

20

21

0.2