|
1
|
|
|
import numpy as np |
|
2
|
|
|
from sklearn.datasets import load_breast_cancer |
|
3
|
|
|
from sklearn.model_selection import cross_val_score |
|
4
|
|
|
from sklearn.decomposition import PCA |
|
5
|
|
|
from sklearn.feature_selection import SelectKBest, f_classif |
|
6
|
|
|
from sklearn.ensemble import GradientBoostingClassifier |
|
7
|
|
|
from hyperactive import Hyperactive |
|
8
|
|
|
|
|
9
|
|
|
data = load_breast_cancer() |
|
10
|
|
|
X, y = data.data, data.target |
|
11
|
|
|
|
|
12
|
|
|
|
|
13
|
|
|
def pca(X): |
|
14
|
|
|
X = PCA(n_components=10).fit_transform(X) |
|
15
|
|
|
|
|
16
|
|
|
return X |
|
17
|
|
|
|
|
18
|
|
|
|
|
19
|
|
|
def none(X): |
|
20
|
|
|
return X |
|
21
|
|
|
|
|
22
|
|
|
|
|
23
|
|
View Code Duplication |
def model(para, X, y): |
|
|
|
|
|
|
24
|
|
|
model = GradientBoostingClassifier( |
|
25
|
|
|
n_estimators=para["n_estimators"], |
|
26
|
|
|
max_depth=para["max_depth"], |
|
27
|
|
|
min_samples_split=para["min_samples_split"], |
|
28
|
|
|
min_samples_leaf=para["min_samples_leaf"], |
|
29
|
|
|
) |
|
30
|
|
|
|
|
31
|
|
|
X_pca = para["decomposition"](X) |
|
32
|
|
|
X = np.hstack((X, X_pca)) |
|
33
|
|
|
|
|
34
|
|
|
X = SelectKBest(f_classif, k=para["k"]).fit_transform(X, y) |
|
35
|
|
|
scores = cross_val_score(model, X, y, cv=3) |
|
36
|
|
|
|
|
37
|
|
|
return scores.mean() |
|
38
|
|
|
|
|
39
|
|
|
|
|
40
|
|
|
search_config = { |
|
41
|
|
|
model: { |
|
42
|
|
|
"decomposition": [pca, none], |
|
43
|
|
|
"k": range(2, 30), |
|
44
|
|
|
"n_estimators": range(10, 200, 10), |
|
45
|
|
|
"max_depth": range(2, 12), |
|
46
|
|
|
"min_samples_split": range(2, 12), |
|
47
|
|
|
"min_samples_leaf": range(1, 11), |
|
48
|
|
|
} |
|
49
|
|
|
} |
|
50
|
|
|
|
|
51
|
|
|
|
|
52
|
|
|
opt = Hyperactive(X, y) |
|
53
|
|
|
opt.search(search_config, n_iter=100) |
|
54
|
|
|
|
SimonBlanke /
Hyperactive
Loading history...