-
Notifications
You must be signed in to change notification settings - Fork 0
/
features_pairs.py
217 lines (166 loc) · 5.19 KB
/
features_pairs.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
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
import data_io
import numpy as np
from sklearn.base import BaseEstimator
from scipy.special import psi
from scipy.stats.stats import pearsonr
from scipy import stats
from scipy.spatial import distance
import copy
import multiprocessing
class FeatureMapper:
def __init__(self, features):
self.features = features
def get_params(self, deep=1):
return {"features": self.features}
def fit(self, X, y=None):
for feature_name, column_names, extractor in self.features:
extractor.fit(X[column_names], y)
def transform(self, X):
extracted = []
for feature_name, column_names, extractor in self.features:
fea = extractor.transform(X[column_names])
if hasattr(fea, "toarray"):
extracted.append(fea.toarray())
else:
extracted.append(fea)
if len(extracted) > 1:
ans = np.concatenate(extracted, axis=1)
else:
ans = extracted[0]
return ans
def fit_transform(self, X, y=None):
extracted = []
for feature_name, column_names, extractor in self.features:
print feature_name
fea = extractor.fit_transform(X[column_names], y)
if hasattr(fea, "toarray"):
extracted.append(fea.toarray())
else:
extracted.append(fea)
if len(extracted) > 1:
ans = np.concatenate(extracted, axis=1)
else:
ans = extracted[0]
return ans
def identity(x):
return x
def rng(x):
return (np.max(x) - np.min(x))
def median(x):
return np.median(x)
def percentile(x, q):
return np.percentile(x, q)
def percentile25(x):
return np.percentile(x, 25)
def percentile75(x):
return np.percentile(x, 75)
def sharpe(x):
v = stats.variation(x)
if np.isinf(v):
return 0
else:
return v
def bollinger(x):
m = np.mean(x)
mx = np.max(x)
mn = np.min(x)
if np.allclose(mx, mn):
return (mx - m)
else:
return ( (mx - m)/(mx - mn) )
def count_unique(x):
return len(set(x))
def normalized_entropy(x):
x = (x - np.mean(x)) / np.std(x)
x = np.sort(x)
hx = 0.0;
for i in range(len(x)-1):
delta = x[i+1] - x[i];
if delta != 0:
hx += np.log(np.abs(delta));
hx = hx / (len(x) - 1) + psi(len(x)) - psi(1);
return hx
def entropy_difference( (x, y) ):
nx = normalized_entropy(x)
ny = normalized_entropy(y)
#return normalized_entropy(x) - normalized_entropy(y)
return [nx, ny, nx - ny]
def correlation( (x, y) ):
#return list(pearsonr(x, y))
return distance.correlation(x, y)
def correlation_magnitude( (x, y) ):
return np.abs(correlation( (x, y) ))
def linregress( (x, y) ):
return stats.linregress(x, y)
def ttest_ind( (x, y) ):
return list(stats.ttest_ind(x, y))
def ttest_rel_t( (x, y) ):
return float(stats.ttest_rel(x, y)[0])
def ttest_rel_p( (x, y) ):
return stats.ttest_rel(x, y)[1]
def ks_2samp( (x, y) ):
return stats.ks_2samp(x, y)
def kruskal( (x, y) ):
return stats.kruskal(x, y)
def bartlett( (x, y) ):
return stats.bartlett(x, y)
def levene( (x, y) ):
return stats.levene(x, y)
def shapiro(x):
return stats.shapiro(x)
def fligner( (x, y) ):
return stats.fligner(x, y)
def mood( (x, y) ):
return stats.mood(x, y)
def oneway( (x, y) ):
return stats.oneway(x, y)
#Distance based measures
def braycurtis( (x, y) ):
return distance.braycurtis(x, y)
def canberra( (x, y) ):
return distance.canberra(x, y)
def chebyshev( (x, y) ):
return distance.chebyshev(x, y)
def cityblock( (x, y) ):
return distance.cityblock(x, y)
def cosine( (x, y) ):
return distance.cosine(x, y)
def hamming( (x, y) ):
return distance.hamming(x, y)
def minkowski( (x, y) ):
ret = []
for p in range(2, 6):
ret += [ distance.minkowski(x, y, p) ]
return ret
def sqeuclidean( (x, y) ):
return distance.sqeuclidean(x, y)
class SimpleTransform(BaseEstimator):
def __init__(self, transformer=identity):
self.transformer = transformer
def fit(self, X, y=None):
return self
def fit_transform(self, X, y=None):
return self.transform(X)
def transform(self, X, y=None):
return_value = np.array([self.transformer(x) for x in X], ndmin=2).T
if return_value.shape[1] == 1:
return return_value
else:
return return_value.T
# Important: Define G_PROCESS_POOL just before using it and after defining all relevant methods.
G_PROCESS_POOL = multiprocessing.Pool()
class MultiColumnTransform(BaseEstimator):
def __init__(self, transformer):
self.transformer = transformer
def fit(self, X, y=None):
return self
def fit_transform(self, X, y=None):
return self.transform(X)
def transform(self, X, y=None):
return_value = G_PROCESS_POOL.map(self.transformer, [x[1] for x in X.iterrows()])
return_value = np.array(return_value, ndmin=2).T
#return_value = np.array([self.transformer(*x[1]) for x in X.iterrows()], ndmin=2).T
if return_value.shape[1] == 1:
return return_value
else:
return return_value.T