|
1
|
|
|
# disables sklearn warnings |
|
2
|
|
|
def warn(*args, **kwargs): |
|
3
|
|
|
pass |
|
4
|
|
|
|
|
5
|
|
|
|
|
6
|
|
|
import warnings |
|
7
|
|
|
|
|
8
|
|
|
warnings.warn = warn |
|
9
|
|
|
|
|
10
|
|
|
import itertools |
|
11
|
|
|
|
|
12
|
|
|
from sklearn.datasets import load_breast_cancer |
|
13
|
|
|
from sklearn.model_selection import cross_val_score |
|
14
|
|
|
from mlxtend.classifier import StackingClassifier |
|
15
|
|
|
|
|
16
|
|
|
from sklearn.ensemble import ( |
|
17
|
|
|
GradientBoostingClassifier, |
|
18
|
|
|
RandomForestClassifier, |
|
19
|
|
|
ExtraTreesClassifier, |
|
20
|
|
|
) |
|
21
|
|
|
|
|
22
|
|
|
from sklearn.neighbors import KNeighborsClassifier |
|
23
|
|
|
from sklearn.neural_network import MLPClassifier |
|
24
|
|
|
from sklearn.gaussian_process import GaussianProcessClassifier |
|
25
|
|
|
from sklearn.tree import DecisionTreeClassifier |
|
26
|
|
|
from sklearn.naive_bayes import GaussianNB |
|
27
|
|
|
|
|
28
|
|
|
from sklearn.linear_model import LogisticRegression |
|
29
|
|
|
from sklearn.linear_model.ridge import RidgeClassifier |
|
30
|
|
|
|
|
31
|
|
|
from hyperactive import Hyperactive |
|
32
|
|
|
|
|
33
|
|
|
data = load_breast_cancer() |
|
34
|
|
|
X, y = data.data, data.target |
|
35
|
|
|
|
|
36
|
|
|
|
|
37
|
|
|
gbc = GradientBoostingClassifier() |
|
38
|
|
|
rfc = RandomForestClassifier() |
|
39
|
|
|
etc = ExtraTreesClassifier() |
|
40
|
|
|
|
|
41
|
|
|
mlp = MLPClassifier() |
|
42
|
|
|
gnb = GaussianNB() |
|
43
|
|
|
gpc = GaussianProcessClassifier() |
|
44
|
|
|
dtc = DecisionTreeClassifier() |
|
45
|
|
|
knn = KNeighborsClassifier() |
|
46
|
|
|
|
|
47
|
|
|
lr = LogisticRegression() |
|
48
|
|
|
rc = RidgeClassifier() |
|
49
|
|
|
|
|
50
|
|
|
|
|
51
|
|
|
def stacking(para, X, y): |
|
52
|
|
|
stack_lvl_0 = StackingClassifier( |
|
53
|
|
|
classifiers=para["lvl_0"], meta_classifier=para["top"] |
|
54
|
|
|
) |
|
55
|
|
|
stack_lvl_1 = StackingClassifier( |
|
56
|
|
|
classifiers=para["lvl_1"], meta_classifier=stack_lvl_0 |
|
57
|
|
|
) |
|
58
|
|
|
scores = cross_val_score(stack_lvl_1, X, y, cv=3) |
|
59
|
|
|
|
|
60
|
|
|
return scores.mean() |
|
61
|
|
|
|
|
62
|
|
|
|
|
63
|
|
|
def get_combinations(models): |
|
64
|
|
|
comb = [] |
|
65
|
|
|
for i in range(0, len(models) + 1): |
|
66
|
|
|
for subset in itertools.permutations(models, i): |
|
67
|
|
|
if len(subset) == 0: |
|
68
|
|
|
continue |
|
69
|
|
|
comb.append(list(subset)) |
|
70
|
|
|
return comb |
|
71
|
|
|
|
|
72
|
|
|
|
|
73
|
|
|
top = [lr, dtc, gnb, rc] |
|
74
|
|
|
models_0 = [gpc, dtc, mlp, gnb, knn] |
|
75
|
|
|
models_1 = [gbc, rfc, etc] |
|
76
|
|
|
|
|
77
|
|
|
stack_lvl_0_clfs = get_combinations(models_0) |
|
78
|
|
|
stack_lvl_1_clfs = get_combinations(models_1) |
|
79
|
|
|
|
|
80
|
|
|
|
|
81
|
|
|
search_config = { |
|
82
|
|
|
stacking: {"lvl_1": stack_lvl_1_clfs, "lvl_0": stack_lvl_0_clfs, "top": top} |
|
83
|
|
|
} |
|
84
|
|
|
|
|
85
|
|
|
|
|
86
|
|
|
opt = Hyperactive(search_config, n_jobs=2, n_iter=150) |
|
87
|
|
|
opt.search(X, y) |
|
88
|
|
|
|
SimonBlanke /
Hyperactive
