Table 4 The details of parameter adjustment

Learning-rate

{0.001,0.002,0.005,0.01,0.02,0.05}

0.001

Learning-rate

{0.001,0.002,0.005,0.01,0.02,0.05}

0.001

Hidden1

{16,32,64,128,256}

256

Hidden1

{16,32,64,128,256}

256

Hidden2

{16,32,64,128,256}

128

Hidden2

{16,32,64,128,256}

128

Dropout

{0,0.1, 0.2,0.4,0.5}

0

Dropout

{0,0.1, 0.2,0.4,0.5}

0.1

Num_neighbors

{1,2,4,5,6,8,10}

5

Num_heads

{1,2,3,4,5,6,7,8,9,10}

2

Learning-rate

{0.001,0.002,0.005,0.01,0.02,0.05}

0.001

Kernel

{'linear', 'poly', 'rbf', 'sigmoid'}

Rbf

Hidden1

{16,32,64,128,256}

256

C

Range(50,150)

100

Hidden2

{16,32,64,128,256}

128

Gamma

{'scale', 'auto'}

Scale

dropout

{0,0.1,0.2,0.4,0.5}

0

n_neighbors

Range(1,50)

8

n_estimators

Range(150,250)

220

Weights

{'uniform', 'distance'}

Distance

Criterion

{"gini", "entropy"}

Gini

Algorithm

{'auto', 'ball_tree', 'kd_tree', 'brute'}

Auto