-
Notifications
You must be signed in to change notification settings - Fork 0
/
ExtractDataset.py
145 lines (118 loc) · 4.35 KB
/
ExtractDataset.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
from statsmodels.tsa.stattools import adfuller
from statsmodels.tsa.arima_model import ARIMA
from sklearn.metrics import mean_squared_error
from math import sqrt
import numpy as np
import pandas as pd
import tsfel
import os
from multiprocessing import Pool
# WandB – Import the wandb library
import wandb
import warnings
warnings.filterwarnings("ignore")
wandb.init(project="meta-arima")
N_CPUS = 1
# Global variable with results
results_dict = {}
# Avaliando o modelo ARIMA com ordens diferentes (p,d,q)
def evaluate_arima_model(X, arima_order):
train_size = int(len(X) * 0.70)
train, test = X[0:train_size], X[train_size:]
history = [x for x in train]
predictions = list()
for t in range(len(test)):
model = ARIMA(history, order=arima_order)
model_fit = model.fit(disp=0)
yhat = model_fit.forecast()[0]
predictions.append(yhat)
history.append(test.iloc[t])
rmse = sqrt(mean_squared_error(test, predictions))
return rmse, arima_order
def evaluate_arima_callback(result):
global results_dict
rmse = result[0]
order = result[1]
wandb.log({"RMSE": rmse})
results_dict[order] = rmse
def evaluate_models(dataset, p_values, d_values, q_values):
global results_dict
results_dict = {}
dataset = dataset.astype('float32')
pool = Pool(processes=N_CPUS)
for p in p_values:
for d in d_values:
for q in q_values:
order = (p, d, q)
try:
pool.apply_async(evaluate_arima_model, (dataset, order),
callback=evaluate_arima_callback)
# rmse = evaluate_arima_model(dataset, order)
# print('ARIMA%s RMSE=%.3f' % (order,rmse))
except Exception:
continue
pool.close()
pool.join()
best_score, best_cfg = float("inf"), None
for order, rmse in results_dict.items():
if rmse < best_score:
best_score, best_cfg = rmse, order
# print('Melhor ARIMA%s RMSE=%.3f' % (best_cfg, best_score))
wandb.log({"Best RMSE": best_score})
wandb.log({"Best p": best_cfg[0]})
wandb.log({"Best d": best_cfg[1]})
wandb.log({"Best q": best_cfg[2]})
return best_cfg, best_score
file_ = "GAP_power_consumption"
url = "GAP_power_consumption.csv"
series = pd.read_csv(url, sep=",", squeeze=True)
series = pd.to_numeric(series, errors='coerce')
series = pd.Series(np.nan_to_num(series))
# trocar para cada conta do COLAB!!! *******************************************
series = series[:100000]
output_file = "DatasetSignalAndTargets_0_100000.csv"
if (os.path.exists(output_file)):
df = pd.read_csv(output_file, index_col=False)
atual = len(df)
else:
atual = 0
# Configuracao para GridSearch
p_values = [0, 1, 2, 4, 6, 8]
d_values = range(0, 3)
q_values = range(0, 3)
#p_values = [0]
#d_values = [0]
#q_values = [0]
cfg = tsfel.get_features_by_domain()
tam = 200
tam_janela = np.round(len(series)/tam)
series_split = np.array_split(series, tam_janela)
dataset = pd.DataFrame()
cont = 0
# wandb.log({"Instance": "ICMC"})
wandb.log({"Instance": "ExtractDatabase"})
for i in series_split[atual:]:
result = adfuller(i)
ADF = pd.DataFrame([result[0]])
ADF_pvalue = pd.DataFrame([result[1]])
if result[1] < 0.05:
cont = cont + 1
andamento = cont/len(series_split)*100
wandb.log({"Window": cont})
wandb.log({"Andamento":andamento})
wandb.log({"ADF_pvalue": result[1]})
best_config, best_RMSE = evaluate_models(i, p_values, d_values, q_values)
best_config = pd.DataFrame([best_config])
best_config.columns = ["p", "d", "q"]
i = i.reset_index(drop=False)
i = pd.DataFrame([i.iloc[:,1]])
features = pd.concat([pd.DataFrame([file_]), #nome do arquivo
pd.DataFrame([atual]), #posicao da janela
pd.DataFrame([best_RMSE]), # RMSE da Config
pd.DataFrame([tam]), #Tamanho da janela
ADF, #resultado do teste ADF
ADF_pvalue, #p-valor do ADF
best_config, # melhor ordem para a ARIMA
i], #todo o sinal
axis=1, ignore_index=False)
features.to_csv(output_file, mode='a', header=False)