-
Notifications
You must be signed in to change notification settings - Fork 3
/
main.py
238 lines (213 loc) · 8.73 KB
/
main.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
225
226
227
228
229
230
231
232
233
234
235
import numpy as np
import pandas as pd
import os
from sklearn.linear_model import Ridge, LogisticRegression
from sklearn.svm import SVR, SVC
from sklearn.neighbors import KNeighborsRegressor, KNeighborsClassifier
from sklearn.model_selection import cross_val_predict, KFold
from models.model import Model
from models.metrics import Meter
from models.visualize import make_figure
from data.aflow_data import AflowData
# %%
def get_models():
# regression grid search parameters
nnr_grid_params = {'n_neighbors': [1]}
ridge_grid_params = {'alpha': np.logspace(-5, 2, 5)}
svr_grid_params = {'C': np.logspace(2, 4, 5),
'gamma': np.logspace(-3, 1, 5)}
# classification grid search parameters
nnc_grid_params = {'n_neighbors': [1]}
logreg_grid_params = {'solver': ['lbfgs'], 'C': np.logspace(-1, 4, 5)}
svc_grid_params = {'C': np.logspace(-1, 4, 5),
'gamma': np.logspace(-2, 2, 5)}
models = {}
# regression models
models['ridge'] = Model(Ridge, ridge_grid_params)
# models['svr'] = Model(SVR, svr_grid_params)
# classification models
models['logreg'] = Model(LogisticRegression,
logreg_grid_params,
classification=True)
# models['svc'] = Model(SVC,
# svc_grid_params,
# classification=True)
# dumb models
models['nnr'] = Model(KNeighborsRegressor, nnr_grid_params)
models['nnc'] = Model(KNeighborsClassifier,
nnc_grid_params,
classification=True)
return models
def eval_model(prop,
gap,
models,
data,
model_type,
holdout_elem,
holdout_struct,
folder):
cv = KFold(n_splits=5, shuffle=True, random_state=1)
X_train_scaled, X_test_scaled = data[0:2]
y_train, y_test = data[2:4]
y_train_labeled, y_test_labeled = data[4:6]
formula_train, formula_test = data[6:8]
train_threshold_x, test_threshold_x = data[8:10]
structure_train, structure_test = data[13:15]
holdout_struct = data[15]
gap_size = data[12]
if model_type == 'ridge_density':
model = models['ridge']
else:
model = models[model_type]
if model.classification:
model.fit(X_train_scaled, y_train_labeled)
y_train_pred = cross_val_predict(model.model,
X_train_scaled,
y_train_labeled,
cv=cv,
method='predict_proba')
y_train_pred = [probability[1] for probability in y_train_pred]
y_train_pred = pd.Series(y_train_pred)
else:
model.fit(X_train_scaled, y_train)
y_train_pred = cross_val_predict(model.model,
X_train_scaled,
y_train,
cv=cv)
y_test_pred = model.predict(X_test_scaled)
y_test_pred_prob = model.predict_proba(X_test_scaled)
model.optimize_threshold(y_train_labeled, y_train_pred)
# save csv files
if holdout_elem is None and holdout_struct is None:
csv_path = 'pred_vs_act_data/'+prop+'/'+str(gap)+'/'
os.makedirs(csv_path, exist_ok=True)
df_csv = pd.DataFrame(y_test)
df_csv['predicted'] = y_test_pred
df_csv.to_csv(csv_path+model_type+'_test.csv')
df_csv = pd.DataFrame(y_train)
df_csv['predicted'] = y_train_pred
df_csv.to_csv(csv_path+model_type+'_train.csv')
make_figure(model.threshold,
y_test,
y_test_pred,
formula_test,
gap_size=gap_size,
test_threshold_x=test_threshold_x,
prop=prop,
gap=gap,
model_type=model_type,
classification=model.classification,
holdout_elem=holdout_elem,
structure=structure_test,
holdout_struct=holdout_struct,
folder=folder)
display_train = False
display_train = True
if display_train:
make_figure(model.threshold,
y_train,
y_train_pred,
formula_train,
gap_size=0,
test_threshold_x=train_threshold_x,
prop=prop,
gap=gap,
model_type=model_type+'train',
classification=model.classification,
holdout_elem=holdout_elem,
structure=structure_train,
holdout_struct=holdout_struct,
folder=folder)
output = [model.threshold,
y_test,
y_test_labeled,
y_test_pred,
y_test_pred_prob]
return output
def main(holdout_elem=None, holdout_structure=None, folder='figures'):
aflow_data = AflowData()
props = ['ael_bulk_modulus_vrh',
'ael_debye_temperature',
'ael_shear_modulus_vrh',
'agl_thermal_conductivity_300K',
'agl_thermal_expansion_300K',
'Egap']
gaps = [0,
4,
8,
12]
model_types = list(get_models().keys())
meter = Meter(props, gaps, model_types)
for prop in props:
for gap in gaps:
models = get_models()
data = aflow_data.get_split(prop,
elem_prop='oliynyk',
gap=gap,
seed_num=10,
holdout_elem=holdout_elem,
holdout_only=False,
holdout_structure=holdout_structure
)
for model_type in model_types:
output = eval_model(prop,
gap,
models,
data,
model_type,
aflow_data.holdout_elem,
holdout_structure,
folder=folder
)
meter.update(prop, gap, model_type, output)
# compare bulk prediction to density rule-of-thumb
if prop == 'ael_bulk_modulus_vrh':
meter.model_types = model_types + ['ridge_density']
data = aflow_data.get_split(prop,
elem_prop='oliynyk',
gap=gap,
seed_num=1,
holdout_elem=holdout_elem,
holdout_only=False,
density_feat=True,
holdout_structure=holdout_structure
)
output = eval_model(prop,
gap,
models,
data,
'ridge_density',
aflow_data.holdout_elem,
holdout_structure,
folder=folder)
meter.update(prop, gap, 'ridge_density', output)
model_types = list(get_models().keys())
return meter
if __name__ == '__main__':
# run with normal train-test split
holdout_elem = None
holdout_structure = None
save_dir = 'figures_default'
meter = main(holdout_elem, holdout_structure, folder=save_dir)
meter.metrics()
meter.plot_curve(curve='roc', folder=save_dir)
meter.plot_curve(curve='pr', folder=save_dir)
meter.save(save_dir)
# remove most common element in 'extraordinary data' from training data
holdout_elem = 0
holdout_structure = None
save_dir = 'figures_missing_elem'
meter = main(holdout_elem, holdout_structure, folder=save_dir)
meter.metrics()
meter.plot_curve(curve='roc', folder=save_dir)
meter.plot_curve(curve='pr', folder=save_dir)
meter.save(save_dir)
# remove most common structure in 'extraordinary data' from training data
holdout_elem = None
holdout_structure = 0
save_dir = 'figures_missing_struct'
meter = main(holdout_elem, holdout_structure, folder=save_dir)
meter.metrics()
meter.plot_curve(curve='roc', folder=save_dir)
meter.plot_curve(curve='pr', folder=save_dir)
meter.save(save_dir)