-
Notifications
You must be signed in to change notification settings - Fork 21
/
utils_ECAD.py
164 lines (147 loc) · 6.18 KB
/
utils_ECAD.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
from keras.optimizers import Adam
from keras.layers import Dropout
from keras.layers import LSTM
from keras.layers import Dense
from sklearn.svm import SVC
from keras.models import Sequential
import sys
import os
import numpy as np
import matplotlib.pyplot as plt
from sklearn.ensemble import RandomForestClassifier
from sklearn.linear_model import LogisticRegression
from PI_class_ECAD import conformal_AD
from sklearn import neighbors
'''Helper Functions '''
def one_dimen_transform(Y_train, Y_predict, d):
n = len(Y_train)
n1 = len(Y_predict)
X_train = np.zeros((n-d, d)) # from d+1,...,n
X_predict = np.zeros((n1, d)) # from n-d,...,n+n1-d
for i in range(n-d):
X_train[i, :] = Y_train[i:i+d]
for i in range(n1):
if i < d:
X_predict[i, :] = np.r_[Y_train[n-d+i:], Y_predict[:i]]
else:
X_predict[i, :] = Y_predict[i-d:i]
Y_train = Y_train[d:]
return([X_train, X_predict, Y_train, Y_predict])
def keras_mod():
model = Sequential(name='NeuralNet')
model.add(Dense(100, activation='relu'))
model.add(Dense(100, activation='relu'))
model.add(Dropout(0.2))
model.add(Dense(100, activation='relu'))
model.add(Dense(1, activation='relu'))
# Compile model
opt = Adam(0.0005)
model.compile(loss='mean_squared_error', optimizer=opt)
return model
def keras_rnn():
model = Sequential(name='RNN')
# For fast cuDNN implementation, activation = 'relu' does not work
model.add(LSTM(100, activation='tanh', return_sequences=True))
model.add(LSTM(100, activation='tanh'))
model.add(Dense(1, activation='relu'))
# Compile model
opt = Adam(0.0005)
model.compile(loss='mean_squared_error', optimizer=opt)
return model
def get_anomalies_classification(data, regr_name, train_size, alpha, stride, dotted,
B_tot=100, rf_num_est=100, phi='mean', neighbor_size=10, return_fitted=False, downsample=True):
# Note, past_memory only used when one_dim=True
# Modeling
# Initialize Parameters
np.random.seed(98765)
B = np.random.binomial(B_tot, np.exp(-1)) # number of bootstrap samples
data_x = data.iloc[:, :-1]
data_y = data.iloc[:, -1]
data_x_numpy = data_x.to_numpy() # Convert to numpy
data_y_numpy = data_y.to_numpy() # Convert to numpy
X_train = data_x_numpy[:train_size, :]
X_predict = data_x_numpy[train_size:, :]
Y_train = data_y_numpy[:train_size]
Y_predict = data_y_numpy[train_size:]
if downsample:
print('Downsample used')
train_abnormal = np.where(Y_train == 1)[0]
train_normal = np.where(Y_train == 0)[0]
X_train_abnormal = X_train[train_abnormal, :]
Y_train_abnormal = Y_train[train_abnormal]
down_sample_idx = np.random.choice(train_normal, 5*len(train_abnormal), replace=False)
X_train_normal = X_train[down_sample_idx, :]
Y_train_normal = Y_train[down_sample_idx]
X_train = np.vstack((X_train_abnormal, X_train_normal))
Y_train = np.hstack((Y_train_abnormal, Y_train_normal))
if regr_name == 'Logistic':
regr = LogisticRegression()
elif regr_name == 'RF':
regr = RandomForestClassifier(n_estimators=rf_num_est, criterion='gini',
bootstrap=False, max_depth=3, n_jobs=-1)
elif regr_name == 'SVC':
regr = SVC(gamma='auto')
elif regr_name == 'kNN':
regr = neighbors.KNeighborsClassifier(n_neighbors=20, weights="distance")
else:
raise ValueError('No specification of regressor')
regr_results = conformal_AD(
regr, X_train, X_predict, Y_train, Y_predict, neighbor_size, phi)
# print('fitting model')
# est_PI, est_nomalies = regr_results.run_experiments(alpha=alpha, B=B, stride=stride, itrial=1)
# if return_fitted:
# print('PIs are returned')
# return(est_PI, est_nomalies)
# else:
# print('PIs are *NOT* returned')
# return(est_nomalies)
est_anomalies = regr_results.run_experiments(
alpha, B, stride, itrial=1, dotted=dotted)
if return_fitted:
return(est_anomalies, regr_results)
return(est_anomalies)
def find_anomalies(p_vals, threshold):
# Note, this allows threshold to be an array or list with length equal to number
# of columns of p_vals
nrow, ncol = p_vals.shape
anomalies = np.where(p_vals.max(axis=0) <= threshold)[0]
return(anomalies)
# Check accuracies
def accuracies(estimate, truth):
precision = sum(np.in1d(estimate, truth))/len(estimate)
recall = sum(np.in1d(estimate, truth))/len(truth)
F1 = 2*(precision*recall)/(precision+recall)
print(f'Precision is {precision}')
print(f'Recall is {recall}')
print(f'F1 score is {F1}')
return(precision, recall, F1)
def plt_prec_recall_F1(df):
"""Plot precision, recall, and F1 score in a row, for all methods
x-axis is the train_frac
"""
fig, ax = plt.subplots(1, 3, figsize=(15, 5), sharex=True, sharey=True)
methods = np.unique(df['method'])
train_frac = np.unique(df['train_frac'])
num_mtds = len(methods)
axisfont = 22
tickfont = axisfont - 2
for j in range(num_mtds):
mtd = methods[j]
df_mtd = df.loc[(df.method == mtd) & (df.itrial == 0)]
precision, recall, F1 = df_mtd['precision'], df_mtd['recall'], df_mtd['F1']
idx_to_name = {0: precision, 1: recall, 2: F1}
idx_to_name_1 = {0: 'Precision', 1: 'Recall', 2: r'$F_1$'}
for i in range(3):
current_fig = ax[i]
current_fig.plot(train_frac, idx_to_name[i], label=mtd, linestyle='-',
marker='o')
# Title, labels, ticks, etc.
current_fig.tick_params(axis='both', which='major', labelsize=tickfont)
current_fig.set_xticks(train_frac)
if i == 0:
current_fig.legend(loc='upper left', bbox_to_anchor=(0.25, 1.45),
ncol=int(num_mtds/2), fontsize=axisfont-2)
current_fig.set_ylabel(idx_to_name_1[i], fontsize=axisfont)
current_fig.set_xlabel('Train Fraction', fontsize=axisfont)
fig.tight_layout(pad=0)
fig.savefig('Kaggle_plts.pdf', dpi=300, bbox_inches='tight', pad_inches=0)