-
Notifications
You must be signed in to change notification settings - Fork 0
/
ModelOptimization.py
224 lines (184 loc) · 6.58 KB
/
ModelOptimization.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
218
219
220
221
222
223
224
from loguru import logger
from sklearn.neural_network import MLPClassifier
from sklearn.svm import SVC
from sklearn.ensemble import RandomForestClassifier
from sklearn.model_selection import StratifiedKFold, GridSearchCV
from sklearn.neighbors import KNeighborsClassifier
from time import time
import pickle
class ModelOptimization:
"""
Class for training and running cross validation on different models
"""
def __init__(self, random_state=None) -> None:
"""
Sets a random state and StratifiedKFold
Parameters:
----------
random_state: provides a seed for all methods that use random_state
"""
self.random_state = random_state
self.skf = StratifiedKFold(
n_splits=5, random_state=self.random_state, shuffle=True
)
def cross_validate(self, X, y, estimator, parameters: dict) -> tuple:
"""
Runs cross validation with the same StratifiedKFold for
all models and train the given model on the whole training data.
Parameters:
----------
X: feature
y: labels
estimator: model
parameters: parameters used in cross validation
return: trained model,
model score,
mean fit time,
mean prediction time,
total cv time
"""
gscv = GridSearchCV(estimator, parameters, cv=self.skf, verbose=3, n_jobs=7)
start_time = time()
gscv.fit(X, y)
end_time = time()
total_validation_time = end_time - start_time
best_model = gscv.best_estimator_
best_score = gscv.best_score_
best_params = gscv.best_params_
best_params_index = gscv.cv_results_["params"].index(best_params)
mean_fit_time = gscv.cv_results_["mean_fit_time"][best_params_index]
mean_pred_time = gscv.cv_results_["mean_score_time"][best_params_index]
print(f"Mean score of best estimator: {best_score}")
print(f"With parameters: {best_params}\n")
print(f"Mean fit time: {mean_fit_time}")
print(f"Mean prediction time: {mean_pred_time}")
print(f"Total cross validation time: {total_validation_time}")
return (
best_model,
best_score,
mean_fit_time,
mean_pred_time,
total_validation_time,
)
def train_and_save_best_models(self, X, y) -> None:
"""
Method for training and saving models
Parameters:
----------
X: features
y: labels
"""
models = dict()
models["SVC"] = self.support_vector(X, y)
models["KNN"] = self.k_neighbors(X, y)
models["RandomForest"] = self.random_forest(X, y)
models["MLPC"] = self.multi_layer_perceptron(X, y)
with open("models/trained_models.pickle", "wb") as f:
pickle.dump(models, file=f)
def load_trained_models(self) -> dict:
"""
Loads trained models from file
Use the model name as key
available models:
SVC - Support Vector Classifier
KNN - K_Nearest Neigbors Classifier
RandomForest - Random Forest Classifier
MLPC - Multi layer perceptron Classifier
return: dictionary of models
"""
with open("models/trained_models.pickle", "rb") as f:
models = pickle.load(f)
return models
def support_vector(self, X, y) -> tuple[SVC, float, float, float]:
"""
Runs cross validation on Suport Vector Machine (SVC)
Parameters:
----------
X: features
y: labels
return: trained model,
model score,
mean fit time,
mean prediction time,
total cv time
"""
parameters = {
"kernel": ["rbf"],
"gamma": [0.1, 0.01, 0.001, 0.0001],
"C": [1, 10, 100, 1000],
}
logger.info("Running parameter testing for support vector")
svc = SVC(random_state=self.random_state)
return self.cross_validate(X, y, svc, parameters)
def random_forest(self, X, y) -> tuple[RandomForestClassifier, float, float, float]:
"""
Runs cross validation on Random Forest Classifier
Parameters:
----------
X: features
y: labels
return: trained model,
model score,
mean fit time,
mean prediction time,
total cv time
"""
parameters = {
"criterion": ["gini", "entropy"],
"max_features": ["sqrt", "log2"],
"min_samples_split": [2, 6, 10],
}
logger.info("Running parameter testing for random forest")
r_forest = RandomForestClassifier(random_state=self.random_state)
return self.cross_validate(X, y, r_forest, parameters)
def k_neighbors(self, X, y) -> tuple[KNeighborsClassifier, float, float, float]:
"""
Runs cross validation on KNearest Neighbors Classifier
Parameters:
----------
X: features
y: labels
return: trained model,
model score,
mean fit time,
mean prediction time,
total cv time
"""
parameters = {
"n_neighbors": list(range(1, 22, 2)),
"weights": ["uniform", "distance"],
"metric": ["manhattan", "minkowski", "euclidean"],
}
logger.info("Running parameter testing for k_neighbors")
knn = KNeighborsClassifier()
return self.cross_validate(X, y, knn, parameters)
def multi_layer_perceptron(self, X, y) -> tuple[MLPClassifier, float, float, float]:
"""
Runs cross validation on Multi Layer Perceptron Classifier (MLPC)
Parameters:
----------
X: features
y: labels
return: trained model,
model score,
mean fit time,
mean prediction time,
total cv time
"""
parameters = {
"hidden_layer_sizes": [
(50,),
(100,),
(200,),
(300,),
(50, 10),
(100, 20),
(200, 40),
(300, 80),
],
"activation": ["tanh", "relu"],
"alpha": [0.0001, 0.001, 0.01],
}
logger.info("Running parameter testing for multi_layer_perceptron")
mlpc = MLPClassifier(random_state=self.random_state, max_iter=800)
return self.cross_validate(X, y, mlpc, parameters)