-
Notifications
You must be signed in to change notification settings - Fork 0
/
rf_model.py
370 lines (292 loc) · 16.5 KB
/
rf_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
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
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
"""
Connor Heaton
Mahony Lab
Random Forest binary classifier
"""
from my_utilities import *
from sklearn.metrics import precision_recall_curve, roc_curve, f1_score
from sklearn.metrics.cluster import adjusted_rand_score
from sklearn.model_selection import train_test_split
from sklearn.ensemble import RandomForestClassifier
from sklearn.cluster import KMeans
from sklearn import metrics
import numpy as np
import shutil
import pickle
import random
import time
import os
class rf_model(object):
def __init__(self, cell_type, raw_features, raw_labels, config, avail_marks, log_path=None, root_dir=None, hist_marks=False, timeit=False, config_black_list=[]):
if timeit:
start_time = time.time()
self.cell_type = cell_type
#self.raw_features = raw_features
self.hist_mark_data = raw_features
self.raw_labels = raw_labels
self.available_mark_names = avail_marks
#self.seq_length = seq_length
"""
self.train_x, self.test_x, self.val_x, self.train_y,\
self.test_y, self.val_y = self.split_data(self.raw_features, self.raw_labels)
print('Feat shape: {} Label shape: {}'.format(self.raw_features.shape, self.raw_labels.shape))
del self.raw_features, self.raw_labels
print('TRAIN feat shape: {} label shape: {}'.format(self.train_x.shape, self.train_y.shape))
print('TEST feat shape: {} label shape: {}'.format(self.test_x.shape, self.test_y.shape))
print('VAL feat shape: {} label shape: {}'.format(self.val_x.shape, self.val_y.shape))
#self.build_model()
"""
if root_dir == None:
self.script_start_time_id = time.strftime('%Y%m%d-%H%M%S')
self.config = dict(config.items())
self.config = read_config('lin_config.txt', self.config, black_list=config_black_list)
self.LOGGING = self.config['logging']
self.root_folder = os.path.join('rfmodels', '{}marks_{}trees_{}'.format(self.config['num_hist_min'], self.config['n_trees'], self.script_start_time_id))
if not os.path.exists(self.root_folder):
os.makedirs(self.root_folder)
write_dict(self.config, os.path.join(self.root_folder, 'config.txt'))
else:
self.root_folder = root_dir
self.script_start_time_id = os.path.basename(self.root_folder)
self.config = dict()
self.config = read_config(os.path.join(self.root_folder, 'config.txt'), self.config)
self.LOGGING = self.config['logging']
if self.LOGGING:
if not os.path.exists(self.root_folder):
os.makedirs(self.root_folder)
if log_path == None:
self.script_log_path = os.path.join(self.root_folder, 'rf_{}_LOG.csv'.format(self.script_start_time_id))
else:
self.script_log_path = log_path
#self.config['model_save_path_tmplt'] =
if self.LOGGING:
if not os.path.exists(os.path.join(self.root_folder, 'models')):
os.makedirs(os.path.join(self.root_folder, 'models'))
if not os.path.exists(os.path.join(self.root_folder, 'preds')):
os.makedirs(os.path.join(self.root_folder, 'preds'))
#self.config['model_save_path_tmplt'] = os.path.join(self.root_folder, 'models', self.config['model_save_path_tmplt'])
#self.config['pred_file_tmplt'] = os.path.join(self.root_folder, 'preds', self.config['pred_file_tmplt'])
self.script_parms = ['cell', 'hist_marks', 'chrom', 'ep_state', 'acc', 'auprc', 'auroc', 'adj_rand_idx', 'f1_score', 'true_in_train', 'tot_train_samples', 'elapsed_time']
print('{}'.format(self.config['pred_file_tmplt']))
"""
if self.config['hist_marks']:
self.hist_write = 'True'
else:
self.hist_write = 'False'
"""
if self.LOGGING:
if not os.path.exists(self.script_log_path):
with open(self.script_log_path, 'a+') as f:
f.write(','.join(self.script_parms) + '\n')
if timeit:
print('Time to init: {0:4.2f}s'.format(time.time() - start_time))
def reset(self):
tf.reset_default_graph()
del self.train_x_state, self.train_y_state
del self.test_x_state, self.test_y_state
def split_data(self, features, labels):
print('Splitting data...')
n = features.shape[0]
train_features, val_features, train_labels, val_labels = train_test_split(features, labels, stratify=labels, test_size=0.1)
train_features, test_features, train_labels, test_labels = train_test_split(train_features, train_labels, stratify=train_labels, test_size=(0.2/0.9))
return train_features, test_features, val_features, train_labels, test_labels, val_labels
def multiple_one_hot_seq(self, cat_int_tensor, depth_list):
"""Creates one-hot-encodings for multiple categorical attributes and
concatenates the resulting encodings
Args:
cat_tensor (tf.Tensor): tensor with mutiple columns containing categorical features
depth_list (list): list of the no. of values (depth) for each categorical
Returns:
one_hot_enc_tensor (tf.Tensor): concatenated one-hot-encodings of cat_tensor
"""
one_hot_enc_tensor = tf.one_hot(cat_int_tensor[:,:,0], depth_list[0], axis=2)
for col in range(1, len(depth_list)):
add = tf.one_hot(cat_int_tensor[:,:,col], depth_list[col], axis=1)
one_hot_enc_tensor = tf.concat([one_hot_enc_tensor, add], axis=2)
return one_hot_enc_tensor
def variable_summaries(self, var):
"""Attach a lot of summaries to a Tensor (for TensorBoard visualization)."""
with self.graph.as_default():
with tf.name_scope('summaries'):
mean = tf.reduce_mean(var)
tf.summary.scalar('mean', mean)
with tf.name_scope('stddev'):
stddev = tf.sqrt(tf.reduce_mean(tf.square(var - mean)))
tf.summary.scalar('stddev', stddev)
tf.summary.scalar('max', tf.reduce_max(var))
tf.summary.scalar('min', tf.reduce_min(var))
tf.summary.histogram('histogram', var)
def build_model(self, timeit=False):
if timeit:
start_time = time.time()
if self.config['pre_cluster']:
self.k_means = KMeans(n_clusters=self.config['n_clusters'], random_state=16, n_jobs=-2, copy_x=True)
self.clf = RandomForestClassifier(n_estimators=100, n_jobs=-1, random_state=16)
if timeit:
print('Time to build model: {0:4.2f}s'.format(time.time() - start_time))
def load_hist_mark_group(self, desired_mark_list):
self.current_hist_marks = desired_mark_list
mark_idx_cols = [self.available_mark_names.index(mark) for mark in desired_mark_list]
self.raw_features = self.hist_mark_data[:, mark_idx_cols]
self.train_x, self.test_x, self.val_x, self.train_y,\
self.test_y, self.val_y = self.split_data(self.raw_features, self.raw_labels)
print('Feat shape: {} Label shape: {}'.format(self.raw_features.shape, self.raw_labels.shape))
del self.raw_features
print('TRAIN feat shape: {} label shape: {}'.format(self.train_x.shape, self.train_y.shape))
print('TEST feat shape: {} label shape: {}'.format(self.test_x.shape, self.test_y.shape))
print('VAL feat shape: {} label shape: {}'.format(self.val_x.shape, self.val_y.shape))
def load_and_sample_for_state(self, state, timeit=False):
if timeit:
start_time = time.time()
self.state = state
train_labels_cat = self.train_y.reshape((-1))
test_labels_cat = self.test_y.reshape((-1))
train_binary_labels = np.zeros((self.train_y.shape[0], 1))
train_binary_labels[train_labels_cat == state, 0] = 1
print('train_binary_labels shape: {}'.format(train_binary_labels.shape))
print('train_binary_labels: {}'.format(train_binary_labels[:6]))
#train_binary_labels[train_labels_cat != state, 0] = 1
test_binary_labels = np.zeros((self.test_y.shape[0], 1))
test_binary_labels[test_labels_cat == state, 0] = 1
#test_binary_labels[test_labels_cat != state, 0] = 1
train_features_state = self.train_x[train_labels_cat == state, :]
train_features_not_state = self.train_x[train_labels_cat != state, :]
test_features_state = self.test_x[test_labels_cat == state, :]
test_features_not_state = self.test_x[test_labels_cat != state, :]
print('Self train x shape: {} train_features_state shape: {}'.format(self.train_x.shape, train_features_state.shape))
#input('...')
train_binary_labels_state = train_binary_labels[train_labels_cat == state, 0]
train_binary_labels_not_state = train_binary_labels[train_labels_cat != state, 0]
test_binary_labels_state = test_binary_labels[test_labels_cat == state, 0]
test_binary_labels_not_state = test_binary_labels[test_labels_cat != state, 0]
num_balance_samples_train = np.minimum(train_binary_labels_state.shape[0], train_binary_labels_not_state.shape[0])
#if num_balance_samples_train >= train_binary_labels_not_state.shape[0]:
# train_features_state = np.append(train_features_state, train_features_not_state, axis=0)
# train_binary_labels_state = np.append(train_binary_labels_state, train_binary_labels_not_state, axis=0)
# test_features_state = np.append(test_features_state, test_features_not_state, axis=0)
# test_binary_labels_state = np.append(test_binary_labels_state, test_binary_labels_not_state, axis=0)
#else:
np.random.seed(16)
train_balance_idx = np.random.choice(train_features_not_state.shape[0], size=num_balance_samples_train, replace=False)
#self.pos_feature_signal, self.neg_feature_signal = get_mean_hist_signal(train_features_state, train_features_not_state[train_balance_idx, :, :], self.root_folder)
train_features_state = np.append(train_features_state, train_features_not_state[train_balance_idx, :], axis=0)
train_binary_labels_state = np.append(train_binary_labels_state, train_binary_labels_not_state[train_balance_idx], axis=0)
print('train_features_state shape: {}'.format(train_features_state.shape))
#input('...')
num_balance_samples_test = np.minimum(test_features_state.shape[0], test_features_not_state.shape[0])
np.random.seed(16)
test_balance_idx = np.random.choice(test_features_not_state.shape[0], size=num_balance_samples_test, replace=False)
test_features_state = np.append(test_features_state, test_features_not_state[test_balance_idx, :], axis=0)
test_binary_labels_state = np.append(test_binary_labels_state, test_binary_labels_not_state[test_balance_idx], axis=0)
del train_binary_labels_not_state, train_features_not_state, test_binary_labels_not_state, test_features_not_state
del train_binary_labels, test_binary_labels
#train_shuffle_idx = np.random.shuffle(np.arange(train_features_state.shape[0]))
#train_shuffle_idx = random.shuffle(list(np.arange(train_features_state.shape[0])))
#print('train_shuffle_idx shape: {}'.format(train_shuffle_idx.shape))
#train_features_state = train_features_state[train_shuffle_idx, :, :]
#train_binary_labels_state = train_binary_labels_state[train_shuffle_idx, :]
c = list(zip(train_features_state, train_binary_labels_state))
random.seed(16)
random.shuffle(c)
train_features_state, train_binary_labels_state = zip(*c)
train_features_state = np.asarray(train_features_state, dtype = np.float32)
train_binary_labels_state = np.asarray(train_binary_labels_state, dtype = np.float32)
states, counts = np.unique(train_binary_labels_state, return_counts=True)
print('States and counts: {}'.format(dict(zip(states, counts))))
#print('train_features_state shape: {}'.format(train_features_state.shape))
#print('train_binary_labels_state: {}'.format(train_binary_labels_state[:6, :]))
#print('test_binary_labels_state: {}'.format(test_binary_labels_state[:6, :]))
#input('...')
#test_shuffle_idx = np.random.shuffle(np.arange(test_features_state.shape[0]))
#test_features_state = test_features_state[test_shuffle_idx, :, :]
#test_binary_labels_state = test_binary_labels_state[test_shuffle_idx, :]
self.train_x_state, self.train_y_state = train_features_state, train_binary_labels_state.reshape(-1)
self.test_x_state, self.test_y_state = test_features_state, test_binary_labels_state.reshape(-1)
print('Creating state directory for state...')
self.state_model_dir = os.path.join(self.root_folder, 'models', '{}'.format(self.cell_type))
self.state_pred_dir = os.path.join(self.root_folder, 'preds', '{}'.format(self.cell_type))
if self.LOGGING:
if not os.path.exists(self.state_model_dir):
os.makedirs(self.state_model_dir)
if not os.path.exists(self.state_pred_dir):
os.makedirs(self.state_pred_dir)
self.state_model_dir = os.path.join(self.root_folder, 'models', '{}'.format(self.cell_type), self.config['model_save_path_tmplt'])
self.state_pred_dir = os.path.join(self.root_folder, 'preds', '{}'.format(self.cell_type), self.config['pred_file_tmplt'])
print('Train x shape, type: {}, {} Train y shape, type: {}, {}'.format(self.train_x_state.shape, self.train_x_state.dtype, self.train_y_state.shape, self.train_y_state.dtype))
if timeit:
print('Time to sample data for state {0}: {1:4.2f}s'.format(self.state, time.time() - start_time))
def calc_auprc(self, binary_labs, pred_prob):
binary_labs = binary_labs.reshape((-1))
pred_prob = pred_prob.reshape((-1))
precision, recall, thresholds = precision_recall_curve(binary_labs, pred_prob)
auprc = metrics.auc(recall, precision)
return auprc
def calc_auroc(self, binary_labs, pred_prob):
binary_labs = binary_labs.reshape((-1))
pred_prob = pred_prob.reshape((-1))
fpr, tpr, thresholds = roc_curve(binary_labs, pred_prob)
auroc = metrics.auc(fpr, tpr)
return auroc
def calc_ari(self, binary_labs, binary_preds):
ari = adjusted_rand_score(binary_labs, binary_preds)
return ari
def calc_f1(self, binary_labs, binary_preds):
f1 = f1_score(binary_labs, binary_preds)
return f1
def train(self):
model_start_time = time.time()
print('Beginning to train model for cell {} state {}...'.format(self.cell_type, self.state))
if self.config['pre_cluster']:
print('Performing k means on train data...')
self.k_means.fit(self.train_x_state)
cluster_labels = self.k_means.labels_
max_x = np.max(cluster_labels) + 1
if self.config['ohe_cluster_labels']:
print('Converting labels to OHE...')
cluster_labels = array_to_one_hot(cluster_labels, max_x)
else:
cluster_labels = np.array(cluster_labels).reshape(-1, 1)
self.train_x_state = np.append(self.train_x_state, cluster_labels, axis=1)
states, counts = np.unique(self.train_y_state, return_counts=True)
print('States and counts: {}'.format(dict(zip(states, counts))))
self.clf.fit(self.train_x_state, self.train_y_state)
if self.config['pre_cluster']:
print('Performing k means on test data...')
cluster_labels = self.k_means.predict(self.test_x_state)
if self.config['ohe_cluster_labels']:
print('Converting labels to OHE...')
cluster_labels = array_to_one_hot(cluster_labels, max_x)
else:
cluster_labels = np.array(cluster_labels).reshape(-1, 1)
self.test_x_state = np.append(self.test_x_state, cluster_labels, axis=1)
print('test_x_state shape: {} test_y_state shape: {}'.format(self.test_x_state.shape, self.test_y_state.shape))
pred_probs = self.clf.predict_proba(self.test_x_state) #, self.test_y)
print('Pred props: {}'.format(pred_probs))
pred_probs = np.array(pred_probs)
print('Pred props post cast: {}'.format(pred_probs))
binary_preds = pred_probs[:,1].reshape(-1)
binary_preds[binary_preds > .5] = 1
binary_preds[binary_preds <= .5] = 0
acc_test = np.mean(self.test_y_state == binary_preds)
auprc = self.calc_auprc(self.test_y_state, pred_probs[:,1])
auroc = self.calc_auroc(self.test_y_state, pred_probs[:,1])
#binary_preds = np.argmax(pred_probs, axis=1).reshape(-1, 1)
f1 = self.calc_f1(self.test_y_state, binary_preds.reshape(-1))
ari = self.calc_ari(self.test_y_state, binary_preds.reshape(-1))
tot_train_samples = self.train_y_state.shape[0]
num_true_in_train = self.train_y_state[self.train_y_state == 1].shape[0]
model_elapsed_time = time.time() - model_start_time
#self.script_parms = ['cell', 'hist_marks', 'chrom', 'ep_state', 'acc', 'auprc', 'auroc', 'adj_rand_idx', 'f1_score', 'true_in_train', 'tot_train_samples', 'elapsed_time']
these_parms = [self.cell_type, '-'.join(self.current_hist_marks), self.config['chrom_id'], self.state, acc_test, auprc, auroc, ari, f1, num_true_in_train, tot_train_samples, model_elapsed_time]
print('Cell: {0} State: {1} Accuracy: {2:3.2f} AUROC: {3:.4f} AUPRC: {4:.4f} Elapsed Time: {5:6.2f}s'.format(self.cell_type, str(self.state), acc_test, auroc, auprc, model_elapsed_time))
print('Saving analysis...')
with open(self.script_log_path, 'a+') as f:
f.write(','.join([str(v) for v in these_parms]) + '\n')
print('Saving predictions...')
# Save predictions and softmax
write_data = np.concatenate([self.test_y_state.reshape(-1, 1), binary_preds.reshape(-1, 1), pred_probs], axis=1)
print('Saving write_data to \'{}\'...'.format(self.config['pred_file_tmplt'].format(self.cell_type, self.state, self.config['hist_marks'])))
np.savetxt(self.state_pred_dir.format(self.cell_type, self.state, self.config['hist_marks']), write_data, delimiter=',')
print('Saving model...')
with open(self.state_model_dir.format(self.cell_type, self.state, '-'.join(self.current_hist_marks)), 'wb') as f:
pickle.dump(self.clf, f)