-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathsimulated_data_generator.py
165 lines (147 loc) · 7.56 KB
/
simulated_data_generator.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
import warnings
import numpy as np
import pandas as pd
from sklearn.model_selection import ShuffleSplit, StratifiedShuffleSplit
class SimulatedDataGenerator(object):
"""This class creates simulation data"""
def __init__(self, sample_shape, missing_portion=0, noise_variance=1, fill_na=np.nan,
positive_correlated=0.25, negative_correlated=0.25, uncorrelated=0.5,
flip_probability=0.0):
assert positive_correlated + negative_correlated + uncorrelated == 1.0
self.sample_shape = sample_shape
self.missing_portion = missing_portion
self.noise_variance = noise_variance
self.fill_na = fill_na
self.data_proportions = [positive_correlated, negative_correlated, uncorrelated]
def generate_data_logistic(self, number_of_samples, proportion_positive=0.5, min_mult=0.2, max_mult=1.0,
flip_probability=0.0):
"""Generates data that would be used for a classification task where the target is either 1 or 0"""
label = np.random.rand(number_of_samples)
flip_vector = np.random.rand(number_of_samples)
flip_vector = flip_vector < flip_probability
flip_vector = flip_vector.astype(int)
y = label < proportion_positive
y = y.astype(int)
x = np.random.normal(loc=0.0, scale=self.noise_variance, size=(number_of_samples, self.sample_shape))
data_edges = np.cumsum(self.data_proportions)
# positively correlated
x[:, :int(data_edges[0] * self.sample_shape)] = x[:, :int(data_edges[0]*self.sample_shape)] + \
np.matmul(np.expand_dims(y, axis=1),
np.abs(np.random.uniform(low=min_mult, high=max_mult,
size=(1, int(self.sample_shape*self.data_proportions[0]))
)
)
)
# negatively correlated
x[:, int(data_edges[0] * self.sample_shape):int(data_edges[1] * self.sample_shape)] = \
x[:, int(data_edges[0] * self.sample_shape):int(data_edges[1] * self.sample_shape)] - \
np.matmul(np.expand_dims(y, axis=1),
np.abs(np.random.uniform(low=min_mult, high=max_mult,
size=(1, int(self.sample_shape*self.data_proportions[1]))
)
)
)
missing_flag = np.zeros((number_of_samples, self.sample_shape))
if self.missing_portion > 0:
missing_probability = np.random.uniform(low=0.0, high=1.0, size=missing_flag.shape)
missing_flag = missing_probability < self.missing_portion
missing_flag = missing_flag.astype(int)
x = np.ma.array(x, mask=missing_flag)
x = x.filled(self.fill_na)
# shuffling samples
sample_order = np.arange(number_of_samples)
np.random.shuffle(sample_order)
x = x[sample_order]
missing_flag = missing_flag[sample_order]
y = y[sample_order]
y = np.abs(y - flip_vector)
return x, missing_flag, y
@staticmethod
def generate_missing(x, missing_portion, fill_na):
"""This generates a version of data already generated with missing values"""
missing_flag = np.zeros(x.shape)
if missing_portion > 0:
missing_probability = np.random.uniform(low=0.0, high=1.0, size=x.shape)
missing_flag = missing_probability < missing_portion
all_missing = x.shape[0] - sum(missing_flag) == 0
if any(all_missing):
missing_flag[0, all_missing] = False
missing_flag = missing_flag.astype(int)
x = np.ma.array(x, mask=missing_flag)
x = x.filled(fill_na)
return x, missing_flag
@staticmethod
def save_data(data, folder_name, split=None):
from os import mkdir, path
x = data[0]
y = data[1]
if len(data) > 2:
missing_flag = data[2]
else:
missing_flag = np.empty(shape=(x.shape[0], 0))
index_column = np.arange(x.shape[0])
column_names = ['var_{}'.format(n) for n in range(x.shape[1])]
if len(data) > 2:
for n in range(missing_flag.shape[1]):
column_names.append('missing_flag_{}'.format(n))
column_names.insert(0, 'ID')
data_concatenated = np.concatenate((np.expand_dims(index_column, axis=1),
x, missing_flag,
np.expand_dims(y, axis=1)), axis=1)
column_names.append('output')
df = pd.DataFrame(data=data_concatenated, columns=column_names)
df.set_index(['ID'], inplace=True)
if split is not None:
df = split_flag(df, split['ratio'], number_of_splits=split['number_of_splits'], stratify=df['output'])
mkdir(folder_name)
to_save_file = path.join(folder_name, 'training_validation_dataset.csv')
df.to_csv(to_save_file)
return df
@staticmethod
def calculate_normalization_values(data, variable_list=None, mode='mean_std', filter_column=None):
"""Normalize numerical variables by either subtracting mean and dividing by standard deviation (mode=mean_std)
or subtracting minimum value and dividing by maximum value (mode=min_max). It returns normalizing values for
saving (even if they are provided in which case they are not changed)"""
if filter_column is not None:
filt = data[filter_column]
else:
filt = data.index
normalizing_values = dict()
if mode == 'mean_std':
for variable in variable_list:
mean_value = data.loc[filt, variable].mean()
std_value = data.loc[filt, variable].std()
normalizing_values[variable] = [mean_value, std_value]
elif mode == 'min_max':
for variable in variable_list:
min_value = data.loc[filt, variable].min()
max_value = data.loc[filt, variable].max()
normalizing_values[variable] = [min_value, max_value - min_value]
else:
warnings.warn('Invalid normalization! Returned raw data')
return normalizing_values
def split_flag(data, ratio, number_of_splits, stratify=None):
"""Create train-test splits and append flags that indicate each one"""
if stratify is None:
split_object = ShuffleSplit(n_splits=number_of_splits,
train_size=ratio,
test_size=1 - ratio,
random_state=0)
all_indices = np.array(data.index)
split_idx = 0
for train_index, test_index in split_object.split(all_indices):
data['train_split_{}'.format(split_idx)] = False
data.loc[data.index[train_index], 'train_split_{}'.format(split_idx)] = True
split_idx += 1
else:
split_object = StratifiedShuffleSplit(n_splits=number_of_splits,
train_size=ratio,
test_size=1 - ratio,
random_state=0)
all_indices = np.array(data.index)
split_idx = 0
for train_index, test_index in split_object.split(all_indices, stratify.values):
data['train_split_{}'.format(split_idx)] = False
data.loc[data.index[train_index], 'train_split_{}'.format(split_idx)] = True
split_idx += 1
return data