-
Notifications
You must be signed in to change notification settings - Fork 7
/
Copy pathtrain.py
168 lines (151 loc) · 6.88 KB
/
train.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
from math import sqrt
from numpy import array2string
from numpy import concatenate
from matplotlib import pyplot
from pandas import DataFrame
from pandas import concat
from sklearn.preprocessing import MinMaxScaler
from sklearn.metrics import mean_squared_error
from keras.models import Sequential
from keras.layers import Dense
from keras.layers import LSTM
from keras.callbacks import EarlyStopping
import argparse
from connectDB import ConnectDB
import time
class trainModel(object):
SYMBOL = 0
ID_COIN = 1
def __init__(self, coin=None, n_hours=1, n_time_predicts=1, config_train=None, units=None):
self.coin = coin
self.n_hours = n_hours
self.n_time_predicts = n_time_predicts
self.config_train = config_train
self.units = units
self.db = ConnectDB()
self.scaler = MinMaxScaler(feature_range=(0, 1))
def series_to_supervised(self, data, n_in=1, n_out=1, dropnan=True):
n_vars = 1 if type(data) is list else data.shape[1]
df = DataFrame(data)
cols, names = list(), list()
for i in range(0, n_out):
cols.append(df.shift(i))
if i == 0:
names += [('var%d(t)' % (j+1)) for j in range(n_vars)]
else:
names += [('var%d(t+%d)' % (j+1, i)) for j in range(n_vars)]
for i in range(n_in, 0, -1):
cols.append(df.shift(-i))
names += [('var%d(t+%d)' % (j+1, i)) for j in range(n_vars)]
agg = concat(cols, axis=1)
agg.columns = names
if dropnan:
agg.dropna(inplace=True)
agg = agg.fillna(0)
return agg
def build_model(self, units, train_X, loss='mse', optimizer='adam'):
model = Sequential()
model.add(LSTM(units,input_shape=(train_X.shape[1], train_X.shape[2]), return_sequences=True))
model.add(LSTM(units))
model.add(Dense(1))
model.compile(loss=loss, optimizer=optimizer)
return model
def save_img_predict_test(self, inv_yhat, inv_y, symbol):
pyplot.plot(inv_yhat, label='predict')
pyplot.plot(inv_y, label='test')
pyplot.legend()
pyplot.savefig("img/chart_%s.png"%symbol)
pyplot.close()
def make_predict(self, model, test_X, n_features = 1):
yhat = model.predict(test_X)
test_X = test_X.reshape((test_X.shape[0], self.n_hours*n_features))
inv_yhat = self.invert_scaling(yhat, test_X, n_features)
return inv_yhat
def make_actual(self, model, test_X, test_y, n_features = 1):
test_X = test_X.reshape((test_X.shape[0], self.n_hours*n_features))
test_y = test_y.reshape((len(test_y), 1))
inv_y = self.invert_scaling(test_y, test_X, n_features)
return inv_y
def invert_scaling(self, test_y, test_X_reshape, n_features):
inv_y = concatenate((test_y, test_X_reshape[:, -(n_features - 1):]), axis=1)
inv_y = self.scaler.inverse_transform(inv_y)
inv_y = inv_y[:,0]
return inv_y
def normalize_data(self, dataset, dropnan=True):
values = dataset.values
values = values.astype('float32')
scaled = self.scaler.fit_transform(values)
reframed = self.series_to_supervised(scaled, self.n_hours, 1, dropnan)
values = reframed.values
return values
def evaluate_model(self, inv_y, inv_yhat):
max_error = 0
for i in range(0,len(inv_y)):
err = abs(inv_y[i] - inv_yhat[i])
if err > max_error:
max_error = err
rmse = sqrt(mean_squared_error(inv_y, inv_yhat))
return max_error, rmse
def split_train_test(self, values, n_time_predicts):
n_train_hours = len(values) - n_time_predicts
train = values[:n_train_hours, :]
test = values[n_train_hours:, :]
return train, test
def split_into_inputs_and_outputs(self, values, n_features = 10):
n_time_predicts = len(values)
n_obs = self.n_hours * n_features
test_X, test_y = values[:, :n_obs], values[:, -n_features]
test_X = test_X.reshape((n_time_predicts, self.n_hours, n_features))
return test_X, test_y
def fit_model(self, model, train_X, train_y, test_X, test_y, symbol, config):
epochs, batch_size, verbose, min_delta, patience, monitor = config
model.fit(train_X, train_y, epochs=epochs, batch_size=batch_size, verbose=verbose, shuffle=False, validation_data=(test_X, test_y),
callbacks = [EarlyStopping(monitor=monitor, min_delta=min_delta, patience=patience)])
return model
def save_model(self, model, symbol):
model.save_weights("weights/weight_%s.h5"%symbol)
model_json = model.to_json()
with open("models/model_%s.json"%symbol, "w") as json_file:
json_file.write(model_json)
def save_to_db(self, inv_y, inv_yhat, id_coin):
max_openTime = self.db.get_max_open_time(id_coin)
max_error, RMSE = self.evaluate_model(inv_y, inv_yhat)
openTime_last = max_openTime - self.n_time_predicts * 60 * 60 * 1000
time_create = int(time.time())
price_predict = array2string(inv_yhat, separator=',')
price_test = array2string(inv_y, separator=',')
self.db.insert_history_train(id_coin, time_create, price_test, price_predict, RMSE, max_error, openTime_last)
def train_model(self):
dataset = self.db.get_data_train_by_id(coin[self.ID_COIN])
n_features = len(dataset.columns)
values = self.normalize_data(dataset)
train, test = self.split_train_test(values, self.n_time_predicts)
train_X, train_y = self.split_into_inputs_and_outputs(train, n_features=n_features)
test_X, test_y = self.split_into_inputs_and_outputs(test, n_features=n_features)
model = self.build_model(units=self.units, train_X = train_X)
model = self.fit_model(model, train_X, train_y, test_X, test_y, coin[self.SYMBOL], self.config_train)
self.save_model(model, coin[self.SYMBOL])
inv_yhat = self.make_predict(model, test_X, n_features)
inv_y = self.make_actual(model, test_X, test_y, n_features)
inv_yhat = inv_yhat[:-1]
inv_y = inv_y[1:]
self.save_to_db(inv_y, inv_yhat, coin[self.ID_COIN])
self.save_img_predict_test(inv_y, inv_yhat, coin[self.SYMBOL])
if __name__ == '__main__':
parser = argparse.ArgumentParser(description="Predict coin price.")
parser.add_argument('-id',type=int,help="-id id coin")
parser.add_argument('-symbol',type=str,help="-symbol symbol coin")
args = parser.parse_args()
coin = [args.symbol, args.id]
units = 64
n_hours = 1
n_time_predicts = 2 * 24
epochs = 50
batch_size = 128
verbose = 0
min_delta = 1e-15
patience = 30
monitor = 'val_loss'
config_train = (epochs, batch_size, verbose, min_delta, patience, monitor)
trainModel = trainModel(coin , n_hours, n_time_predicts, config_train, units)
trainModel.train_model()