-
Notifications
You must be signed in to change notification settings - Fork 0
/
imputation.py
159 lines (127 loc) · 6.24 KB
/
imputation.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
import torch
import torch.nn as nn
import pandas as pd
import random
import numpy as np
import json
from sklearn.preprocessing import StandardScaler
import matplotlib.pyplot as plt
import os
from sklearn.metrics import mean_squared_error as MSE
from sklearn.metrics import mean_absolute_error as MAE
from pypots.imputation import SAITS,BRITS,GPVAE,MRNN,USGAN
TF_ENABLE_ONEDNN_OPTS=0
import argparse
from utils import *
def random_subarray(array, period=24):
array1 = np.copy(array)
array2 = np.copy(array)
length = len(array)
num_to_select = int(length * 0.4)
selected_length = 0
replaced_indices = []
replaced_intervals = []
while selected_length < num_to_select:
period_lengths = [period // 10, period // 4, period // 5, period // 2, period, period * 2, period * 4, period * 5]
subarray_length = random.choice(period_lengths)
if subarray_length + selected_length > num_to_select:
subarray_length = num_to_select - selected_length
start_idx = random.randint(0, length - subarray_length)
end_idx = start_idx + subarray_length - 1
overlapping = False
for interval in replaced_intervals:
if (start_idx >= interval[0] and start_idx <= interval[1]) or (end_idx >= interval[0] and end_idx <= interval[1]):
overlapping = True
break
if overlapping:
continue
array1[start_idx:end_idx + 1] = np.mean(array[start_idx:end_idx + 1])
array2[start_idx:end_idx + 1] = np.nan
replaced_indices.extend(range(start_idx, end_idx + 1))
replaced_intervals.append((start_idx, end_idx))
selected_length += subarray_length
return array1, array2, replaced_indices
def replace_nan_with_array(array1, array2):
if array1.shape != array2.shape:
raise ValueError("error")
result_array = np.copy(array1)
nan_indices = np.isnan(array1)
result_array[nan_indices] = array2[nan_indices]
return result_array
def main(args):
setup_seed(0)
name = args.name
repeat_times = args.repeat_times
device = args.device
data = pd.read_csv(args.data_path,index_col=0)
data.index = pd.to_datetime(data.index)
# deine timestep
time_1 = pd.Timestamp(args.time_1)
time_2 = pd.Timestamp(args.time_2)
time_3 = pd.Timestamp(args.time_3)
# split the data
train = np.array(data.loc[time_1:time_2].iloc[:,-1:]).reshape(-1)
val = np.array(data.loc[time_2:time_3].iloc[:,-1:])
test = np.array(data.loc[time_3:].iloc[:,-1:])
train_mean,train_nan,nan_index = random_subarray(train,24)
train_mean = train_mean.reshape(-1,1)
train_nan = train_nan.reshape(-1,1)
scaler = StandardScaler()
scaler.fit(train_mean)
train_mean = scaler.transform(train_mean)
train_nan = scaler.transform(train_nan)
val = scaler.transform(val)
test = scaler.transform(test)
train_data_mean,train_label_mean = generate_data(train_mean)
train_data_nan,train_label_nan = generate_data(train_nan)
val_data,val_label = generate_data(val)
test_data,test_label = generate_data(test)
create_directory_if_not_exists('./imputated_folder')
create_directory_if_not_exists('./imputated_folder/'+name)
np.save('./imputated_folder/'+name+'/train_data_mean.npy', train_data_mean)
np.save('./imputated_folder/'+name+'/train_label_mean.npy', train_label_mean)
np.save('./imputated_folder/'+name+'/val_data.npy', val_data)
np.save('./imputated_folder/'+name+'/val_label.npy', val_label)
np.save('./imputated_folder/'+name+'/test_data.npy', test_data)
np.save('./imputated_folder/'+name+'/test_label.npy', test_label)
im_dataset = {"X": train_label_nan}
for i in range(repeat_times):
setup_seed(i+1)
model = SAITS(n_steps=24, n_features=1, n_layers=2, d_model=64, d_ffn=32, n_heads=4, d_k=16, d_v=16, dropout=0.1, epochs=100,patience=10,device = device)
model.fit(im_dataset)
imputation = model.impute(im_dataset)
im_result = replace_nan_with_array(train_label_nan,imputation)
np.save('./imputated_folder/'+name+'/train_label_saits_'+str(i+1)+'_.npy',im_result)
model = BRITS(n_steps=24, n_features=1, rnn_hidden_size=64,epochs=100,patience=10,device = device)
model.fit(im_dataset)
imputation = model.impute(im_dataset)
im_result = replace_nan_with_array(train_label_nan,imputation)
np.save('./imputated_folder/'+name+'/train_label_brits_'+str(i+1)+'_.npy',im_result)
model = GPVAE(n_steps=24, n_features=1, latent_size=64,epochs=100,patience=10,device = device)
model.fit(im_dataset)
imputation = model.impute(im_dataset)[:,0,:,:]
im_result = replace_nan_with_array(train_label_nan,imputation)
np.save('./imputated_folder/'+name+'/train_label_gpvae_'+str(i+1)+'_.npy',im_result)
if name != 'ELE':
model = MRNN(n_steps=24, n_features=1, rnn_hidden_size=64,epochs=100,patience=10,device = device)
model.fit(im_dataset)
imputation = model.impute(im_dataset)
im_result = replace_nan_with_array(train_label_nan,imputation)
np.save('./imputated_folder/'+name+'/train_label_mrnn_'+str(i+1)+'_.npy',im_result)
model = USGAN(n_steps=24, n_features=1, rnn_hidden_size=64,epochs=100,patience=10,device = device)
model.fit(im_dataset)
imputation = model.impute(im_dataset)
im_result = replace_nan_with_array(train_label_nan,imputation)
np.save('./imputated_folder/'+name+'/train_label_usgan_'+str(i+1)+'_.npy',im_result)
if __name__ == '__main__':
parser = argparse.ArgumentParser()
parser.add_argument("-p","--data_path", help="data path",default='./data/GEF.csv')
parser.add_argument("-n","--name", help="name of the data",default='GEF',choices=['GEF','ETTH1','ETTH2','ELE','Traffic','Air'])
parser.add_argument("-rt","--repeat_times", type=int,help="times of repeating experiment",default=3)
parser.add_argument("-d","--device",help="device to run",default='cuda')
args = parser.parse_args()
with open("./time_settings.json", "r") as file:
time_settings = json.load(file)
if args.name in time_settings:
args.time_1, args.time_2, args.time_3 = time_settings[args.name]
main(args)