|
1
|
|
|
import numpy as np |
|
2
|
|
|
from sklearn.model_selection import cross_val_score |
|
3
|
|
|
from sklearn.ensemble import GradientBoostingClassifier |
|
4
|
|
|
from sklearn.ensemble import RandomForestClassifier |
|
5
|
|
|
from sklearn.ensemble import ExtraTreesClassifier |
|
6
|
|
|
from sklearn.datasets import load_breast_cancer |
|
7
|
|
|
from hyperactive import Hyperactive |
|
8
|
|
|
|
|
9
|
|
|
data = load_breast_cancer() |
|
10
|
|
|
X, y = data.data, data.target |
|
11
|
|
|
|
|
12
|
|
|
|
|
13
|
|
View Code Duplication |
def model0(para, X, y): |
|
|
|
|
|
|
14
|
|
|
etc = ExtraTreesClassifier( |
|
15
|
|
|
n_estimators=para["n_estimators"], |
|
16
|
|
|
criterion=para["criterion"], |
|
17
|
|
|
max_features=para["max_features"], |
|
18
|
|
|
min_samples_split=para["min_samples_split"], |
|
19
|
|
|
min_samples_leaf=para["min_samples_leaf"], |
|
20
|
|
|
bootstrap=para["bootstrap"], |
|
21
|
|
|
) |
|
22
|
|
|
scores = cross_val_score(etc, X, y, cv=3) |
|
23
|
|
|
|
|
24
|
|
|
return scores.mean() |
|
25
|
|
|
|
|
26
|
|
|
|
|
27
|
|
View Code Duplication |
def model1(para, X, y): |
|
|
|
|
|
|
28
|
|
|
rfc = RandomForestClassifier( |
|
29
|
|
|
n_estimators=para["n_estimators"], |
|
30
|
|
|
criterion=para["criterion"], |
|
31
|
|
|
max_features=para["max_features"], |
|
32
|
|
|
min_samples_split=para["min_samples_split"], |
|
33
|
|
|
min_samples_leaf=para["min_samples_leaf"], |
|
34
|
|
|
bootstrap=para["bootstrap"], |
|
35
|
|
|
) |
|
36
|
|
|
scores = cross_val_score(rfc, X, y, cv=3) |
|
37
|
|
|
|
|
38
|
|
|
return scores.mean() |
|
39
|
|
|
|
|
40
|
|
|
|
|
41
|
|
View Code Duplication |
def model2(para, X, y): |
|
|
|
|
|
|
42
|
|
|
gbc = GradientBoostingClassifier( |
|
43
|
|
|
n_estimators=para["n_estimators"], |
|
44
|
|
|
learning_rate=para["learning_rate"], |
|
45
|
|
|
max_depth=para["max_depth"], |
|
46
|
|
|
min_samples_split=para["min_samples_split"], |
|
47
|
|
|
min_samples_leaf=para["min_samples_leaf"], |
|
48
|
|
|
subsample=para["subsample"], |
|
49
|
|
|
max_features=para["max_features"], |
|
50
|
|
|
) |
|
51
|
|
|
scores = cross_val_score(gbc, X, y, cv=3) |
|
52
|
|
|
|
|
53
|
|
|
return scores.mean() |
|
54
|
|
|
|
|
55
|
|
|
|
|
56
|
|
|
search_config = { |
|
57
|
|
|
model0: { |
|
58
|
|
|
"n_estimators": range(10, 200, 10), |
|
59
|
|
|
"criterion": ["gini", "entropy"], |
|
60
|
|
|
"max_features": np.arange(0.05, 1.01, 0.05), |
|
61
|
|
|
"min_samples_split": range(2, 21), |
|
62
|
|
|
"min_samples_leaf": range(1, 21), |
|
63
|
|
|
"bootstrap": [True, False], |
|
64
|
|
|
}, |
|
65
|
|
|
model1: { |
|
66
|
|
|
"n_estimators": range(10, 200, 10), |
|
67
|
|
|
"criterion": ["gini", "entropy"], |
|
68
|
|
|
"max_features": np.arange(0.05, 1.01, 0.05), |
|
69
|
|
|
"min_samples_split": range(2, 21), |
|
70
|
|
|
"min_samples_leaf": range(1, 21), |
|
71
|
|
|
"bootstrap": [True, False], |
|
72
|
|
|
}, |
|
73
|
|
|
model2: { |
|
74
|
|
|
"n_estimators": range(10, 200, 10), |
|
75
|
|
|
"learning_rate": [1e-3, 1e-2, 1e-1, 0.5, 1.0], |
|
76
|
|
|
"max_depth": range(1, 11), |
|
77
|
|
|
"min_samples_split": range(2, 21), |
|
78
|
|
|
"min_samples_leaf": range(1, 21), |
|
79
|
|
|
"subsample": np.arange(0.05, 1.01, 0.05), |
|
80
|
|
|
"max_features": np.arange(0.05, 1.01, 0.05), |
|
81
|
|
|
}, |
|
82
|
|
|
} |
|
83
|
|
|
|
|
84
|
|
|
|
|
85
|
|
|
opt = Hyperactive(X, y) |
|
86
|
|
|
opt.search(search_config, n_iter=30, n_jobs=4) |
|
87
|
|
|
|
SimonBlanke /
Hyperactive
Loading history...