forked from h2oai/driverlessai-recipes
-
Notifications
You must be signed in to change notification settings - Fork 1
/
pca_transformer.py
51 lines (37 loc) · 1.41 KB
/
pca_transformer.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
"""Principal Component Analysis (PCA) Transformer"""
from h2oaicore.transformer_utils import CustomTransformer
import datatable as dt
from typing import List
class PrincipalComponentAnalysisTransformer(CustomTransformer):
_unsupervised = True
_unsupervised = True
_display_name = "Principal Component Analysis (PCA) Transformer"
@staticmethod
def get_default_properties():
return dict(
col_type="numeric", min_cols=2, max_cols="all", relative_importance=1
)
@staticmethod
def get_parameter_choices():
return dict(n_components=[1, 2, 3])
def __init__(self, n_components=1, **kwargs):
super().__init__(**kwargs)
self._n_components = n_components
def fit_transform(self, X, y=None, **fit_params):
from sklearn.decomposition import PCA
from sklearn.impute import SimpleImputer
X = X.to_numpy()
imp = SimpleImputer()
X = imp.fit_transform(X)
n_components = self._n_components
if min(X.shape) <= n_components:
n_components = min(X.shape) - 1
self.pca = PCA(n_components=n_components)
self.pca.fit(X)
return self.pca.transform(X)
def transform(self, X, y=None, **fit_params):
from sklearn.impute import SimpleImputer
X = X.to_numpy()
imp = SimpleImputer()
X = imp.fit_transform(X)
return self.pca.transform(X)