|
1
|
|
|
# Author: Simon Blanke |
|
2
|
|
|
# Email: [email protected] |
|
3
|
|
|
# License: MIT License |
|
4
|
|
|
|
|
5
|
|
|
import pandas as pd |
|
6
|
|
|
|
|
7
|
|
|
from sklearn.ensemble import GradientBoostingRegressor |
|
8
|
|
|
from sklearn.externals import joblib |
|
9
|
|
|
from sklearn.preprocessing import MinMaxScaler |
|
10
|
|
|
|
|
11
|
|
|
from ._recognizer import Recognizer |
|
12
|
|
|
from ._predictor import Predictor |
|
13
|
|
|
|
|
14
|
|
|
|
|
15
|
|
|
class MetaRegressor: |
|
16
|
|
|
def __init__(self): |
|
17
|
|
|
self.meta_reg = None |
|
18
|
|
|
self.score_col_name = "mean_test_score" |
|
19
|
|
|
|
|
20
|
|
|
def fit(self, X, y): |
|
21
|
|
|
self._train_regressor(X, y) |
|
22
|
|
|
|
|
23
|
|
|
def predict(self, X, y): |
|
24
|
|
|
self.recognizer = Recognizer(self.search_config) |
|
25
|
|
|
self.predictor = Predictor(self.search_config, self.meta_regressor_path) |
|
26
|
|
|
|
|
27
|
|
|
X_test = self.recognizer.get_test_metadata([X, y]) |
|
28
|
|
|
|
|
29
|
|
|
best_hyperpara_dict, best_score = self.predictor.search(X_test) |
|
30
|
|
|
|
|
31
|
|
|
return best_hyperpara_dict, best_score |
|
32
|
|
|
|
|
33
|
|
|
def _scale(self, y): |
|
34
|
|
|
# scale the score -> important for comparison of meta data from datasets in meta regressor training |
|
35
|
|
|
scaler = MinMaxScaler() |
|
36
|
|
|
y = scaler.fit_transform(y) |
|
37
|
|
|
y = pd.DataFrame(y, columns=["mean_test_score"]) |
|
38
|
|
|
|
|
39
|
|
|
return y |
|
40
|
|
|
|
|
41
|
|
|
def _train_regressor(self, X, y): |
|
42
|
|
|
if self.meta_reg is None: |
|
43
|
|
|
n_estimators = int(y.shape[0] / 50 + 50) |
|
44
|
|
|
|
|
45
|
|
|
self.meta_reg = GradientBoostingRegressor(n_estimators=n_estimators) |
|
46
|
|
|
self.meta_reg.fit(X, y) |
|
47
|
|
|
|
|
48
|
|
|
def store_model(self, path): |
|
49
|
|
|
joblib.dump(self.meta_reg, path) |
|
50
|
|
|
|
|
51
|
|
|
def load_model(self, path): |
|
52
|
|
|
self.meta_reg = joblib.load(path) |
|
53
|
|
|
|
SimonBlanke /
Hyperactive
