-
Notifications
You must be signed in to change notification settings - Fork 3
/
model.py
70 lines (52 loc) · 2.26 KB
/
model.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
from .utils import toseries
from numpy import r_, array
from thunder.series import fromrdd
class MassRegressionModel:
"""
A fitted mass univariate regression model.
Contains a collection regression models, each fitted with the same design
matrix, but to a different response varaible.
"""
def __init__(self, models):
self.models = models
@property
def coef_(self):
return self.models.map(lambda v: v[0].coef_)
@property
def intercept_(self):
return self.models.map(lambda v: v[0].intercept_)
@property
def betas(self):
def getbetas(model):
return r_[model.intercept_, model.coef_]
return self.models.map(lambda v: getbetas(v[0]))
def predict(self, X):
return self.models.map(lambda v: v[0].predict(X))
def score(self, X, y):
y = toseries(y)
if y.mode == "spark":
if not self.models.mode == "spark":
raise ValueError("model is spark mode, input y must also be spark mode")
joined = self.models.tordd().join(y.tordd())
result = joined.mapValues(lambda v: array([v[0][0].score(X, v[1])]))
return fromrdd(result)
if y.mode == "local":
if not self.models.mode == "local":
raise ValueError("mode is local mode, input y must also be local mode")
return self.models.map(lambda kv: kv[1][0].score(X, y.values[kv[0]]), with_keys=True)
def predict_and_score(self, X, y):
y = toseries(y)
def get_both(model, X, y):
return r_[model.score(X, y), model.predict(X)]
if y.mode == "spark":
if not self.models.mode == "spark":
raise ValueError("model is spark mode, input y must also be spark mode")
joined = self.models.tordd().join(y.tordd())
both = fromrdd(joined.mapValues(lambda v: get_both(v[0][0], X, v[1])))
if y.mode == "local":
if not self.models.mode == "local":
raise ValueError("mode is local mode, input y must also be local mode")
both = self.models.map(lambda kv: get_both(kv[1][0], X, y.values[kv[0]]), with_keys=True)
predictions = both[:, 1:]
scores = both[:, [0]]
return predictions, scores