-
Notifications
You must be signed in to change notification settings - Fork 94
/
quantile_winsorizer.py
40 lines (31 loc) · 1.26 KB
/
quantile_winsorizer.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
"""Winsorizes (truncates) univariate outliers outside of a given quantile threshold"""
from h2oaicore.transformer_utils import CustomTransformer
import datatable as dt
import numpy as np
class MyQuantileWinsorizer(CustomTransformer):
_unsupervised = True
_testing_can_skip_failure = False # ensure tested as if shouldn't fail
@staticmethod
def get_default_properties():
return dict(col_type="numeric", min_cols=1, max_cols=1, relative_importance=1)
@staticmethod
def get_parameter_choices():
return {"quantile": [0.01, 0.001, 0.05]}
@property
def display_name(self):
return "MyQuantileWinsorizer%s" % str(self._quantile)
def __init__(self, quantile=0.01, **kwargs):
super().__init__(**kwargs)
self._quantile = min(quantile, 1 - quantile)
self._lo = None
self._hi = None
def fit_transform(self, X: dt.Frame, y: np.array = None):
vals = X.to_numpy()
self._lo = float(np.quantile(vals, self._quantile))
self._hi = float(np.quantile(vals, 1 - self._quantile))
return self.transform(X)
def transform(self, X: dt.Frame):
X = dt.Frame(X)
X[self._lo > dt.f[0], float] = self._lo
X[self._hi < dt.f[0], float] = self._hi
return X